ULTIMATE STRENGTH OF BOX GIRDERS BY FINITE ELEMENT METHOD

by

Cetin Yilmaz

FRITZ ENGINEERING
LABORATORY LIBRARY

A Dissertation

Presented to the Graduate Committee

of Lehigh University
in Candidacy for the Degree of
Doctor of Philosophy
in
Civil Engineering

1975
-..

ACKNOHLEDGMENTS

The author is deeply indebted to Professor Ben T. Yen,

Professor in charge of this dissertation for his guidance and contin-

uous encouragement in the preparation of this thesis. The guidance

of Professors Roger G. Slutter, Chairman, Celal N. Kostem, George

C. Sih and David A. VanHorn, members of the special committee

directing the author's doctoral work, is gratefully acknowledged.

Sincere appreciation is expressed to all the author's

associates in Fritz Engineering Laboratory. Two of the author's

friends need to be mentioned specifically. Dr. Suresh Desai pro-

vided the starting point for the computer program developed in this

study and Dr. Dirk P. duPlessis for the frequent discussions.

Special thanks are extended to Mrs. Dorothy Fielding for

typing the manuscript with great care.

iii
 TABLE OF CONTENTS

Page
ABSTRACT 1

1. INTRODUCTION 2
1.1 Background 2
1.2 Objectives and Scope 5

2. ELASTIC ANALYSIS OF BOX GIRDERS 7

2.1 Problem Formulation and Solution 7
2.1.1 Discretization of the Continuum 7
2.1.2 Evaluation of-Element Properties 8
Plate Bending 10
In-Plane Behavior 12
Superposition of In-Plane and Plate ·
Bending Behavior 13
2.1.3 Assembly and Solution 14
2.2 Evaluation of Formulation 15
2.2.1 ~vo Degrees of Freedom versus Five Degrees
of Freedom 15
2.2.2 Mesh Division Along the Length 18
2.2.3 Mesh Division Across the Webs and Flanges 20
2.3 Comparison of Results with Existing Solutions 22
2.3.1 Multicell Box Girder 22
2.3.2 Wide Flange Beam 25
2.3.3 Thin Flange Deck 26
2.3.4 Composite Box Girder 28
2.4 Examples of Applications 33
2.4.1 Diaphragm Rigidity 33
2.4.2 Diaphragm Spacing 34
2.4.3 Web Slenderness and Concrete Flange 36
Rigidity

3. ULTIMATE STRENGTH 38
3.1 Nonlinear Behavior 38
3.2 Material Properties 40

iv
 TABLE OF CONTENTS (continued)
Page
3.2.1 Stress-Strain Relationship 40
3.2.2 Yield Criteria 42
3.2.3 Subsequent Yielding and Flow Rule 43
3.3 Solution Technique 49
3.3.1 Incremental Method 49
3.3.2 The Tangent Stiffness Method so
3.4 Illustrative Problems 53
3.4.1 Wide Flange Beam 53
3.4;2 Steel Box Girders 56
3.4.3 Composite Box Girders

4. SUMMARY AND CONCLUSIONS 65

TABLES 69

FIGURES 76

REFERENCES 116

APPENDIX - NOMENCLATURE 122

VITA 125

v
 LIST OF TABLES

Table Page

2.1 In-Plane Stiffness Matrix 69

2.2 Comparisons of Two Degree of Freedom and Five 70

Degree of Freedom Results

2.3 Effect of Mesh Division Along the Length 71

2.4 Effect of Divisions Across the Flanges 72

2.5 Effect of Number of Divisions Across the Web 73

2.6 Effect of Element Thickness 74

2.7 Web Slenderness 75

vi
 LIST OF FIGURES

Figure Page
2.1 Finite Element Model and Element Coordinate Axes 76

Box Girder Specimen and Mesh Divisions 77

2.3 Wide Flange Beam 78

2.4 Comparison of Deflections from Two and Five Degree 79

Freedom Formulation

2~5 Normal Stress Distribution for Problem 2 80

(Cross-section at SO-in. from Left Support)

Aspect Ratio and Mesh Division along the Length 81

2.7 Effect of Mesh Division along the Length under 82

Simple Bending

2.8 Effect of Mesh Division along the Length under 83

Torsional Loading

2.9 Effect of Mesh Division on Accuracy and Computational 84

Time

2.10 Two Span Multicell Box Girder, Cross-section and 85

Mesh Sizes

2.11 Vertical Deflection along the Longitudinal Line at 86

the Top of the Loaded Web

2ol2 Vertical Deflection of Bottom Flange at Transverse 87

Sections

2ol3 Longitudinal Distribution of Normal Stress 88

(a ) along a Line in Top Flange
n
2.14 Stress and Deflection of W8 X 31 89

2.15 Cross-section and Finite Element of a Thin Flange 90

2.16 Top Flange Normal Stresses at a Transverse Section 91

2.17 Effect of Two Degree versus Five Degree Freedom 92

Formulation on a Thin Flange ·

vii
 Page

2.18 Dimensions and Geometry of Girder Dl 93

2.19 Deflection along the Length of Girder Dl 94

2.20a Normal Stress Distribution due to a Torsional Loading 95

on Girder Dl (86.25 in. from Left Support)

2.20b Normal Stress Distribution due to Simple Bending 96

of Girder Dl (56.25 in. from Left Support)

2.21 Effect of Diaphragm Thickness on Stresses and 97

Deflection under Pure Torsional Loading

2.22 Principal Stresses in a Diaphragm 98

2.23 Effect of Diaphragm Spacing on Deflection, 99

Load at Midspan

2.24 Effect of Diaphragm Spacing on Deflection, Load 100

Load at Midspan, Exaggerated Scale

2.25 Effect of Diaphragm Spacing on Deflection, 101

Load at Quarter Points

2.26 Effect of Diaphragm Spacing on Deflection, 102

Load at Quarter Points, Exaggerated Scale

3.1 Stress-Strain Relationship for Steel and Concrete 103

3.2 Yield Criteria for Steel and Concrete 104

3.3 Graphical Representation of Incremental Tangent 105

Stiffness Procedure

3.4 Loading and Finite Element Discretization of 106

a Steel I-Beam

3.5 Load Deflection Curve for a Steel I-Beam 107

3.6 Dimensions and Geometry of Specimens Ml and N2 108

3.7 Load versus Midspan Deflection- Specimen Ml 109

3.8 Load versus Shear Stress at Centerline of Panel 5 - 110

Specimen M2

viii
 Page

3.9 Load versus Midspan Deflection of Girder Dl 111

3.10 Load versus Midspan Deflection of Girder D2 112

3.11 Load versus Normal Stress at 3.75 in. away from 113
Midspan at the Bottom Flange

3.12 Principal Stresses in the Web Adjacent to Loading 114

Point

3.13 Yield Zones of the Web of Girder Dl under Different 115

Load Magnitudes

ix
 ABSTRACT

This dissertation presents the results of an analysis of the

behavior of straight rectangular box girders to failure when the loads

are applied either through diaphragms or from the web flange junction.

In the analysis the finite element method is used and material nonlin-

earities are considered.

The basic formulation was developed for the elastic analysis

using five degrees of freedom per nodal point and rectangular elements.

A number of examples of thin walled structures showed that a two degree

of freedom formulation gave results comparable to those of a five

degree of freedom formulation but \~ith relatively less cost. The

method was used for the analysis of four different examples. The

boundary conditions of box girders were closely modeled and shear de-

formation and shear lag effect were automatically considered. The re-

sults were compared with the available test data.

A two degree of freedom formulation was adopted in the non-

linear analysis, so that the application of two dimensional failure

criterion was possible. The incremental tangent stiffness method and

the incremental theory of plasticity were selected for the solution

of the nonlinear problem. The load deflection behavior of-some tested

box girders was closely predicted when the geometric nonlinearities

were small. Ultimate load carrying capacity was obtained and was

compared with test results. The method also provided an accurate

means of analysis for the stresses and deflections in both the elastic

range and the inelastic range of material behavior.

-1-
 1. INTRODUCTION

1.1 Background

The use of thin walled steel box girders and steel composite

box girders as main load carrying members in bridge structures has

gained considerable popularity in the last two decadeso Box girders

are designed to function three-aimensionally. Overlapping functions

of the deck plates, stringers, and floor beams are replaced by the

ability of the box to resist bending, shear and torsional loading.

It is this structural efficiency of box girders that leads to their

relative economy in construction and usage as bridge members (1, 2).

As a result of recent box girder bridge failures during con-

struction (3), research activities have been dramatically increased,

particularly in Great Britain and in European countries (4, 5, 6, 7).

Most of these research efforts are directed to nonlinear analysis of

component parts of steel box girders. These studies complement the

numerous available methods of analyzing box shapes. The concepts of

design have been that of preventing buckling of the compon~nt parts

and of analyzing box shapes with due consideration to warping and

distortion of the cross section in the elastic range of material

behavior.

The status of box girder design and analysis has been re-

viewed a number of times in recent years (1966-1974) (1, 2, 8). Of

-2-
the available procedures, the one which is based on the prismatic

folded plate theory of Goldberg and Leve (9) offers an accurate method

of analysis. This method considers box girders as a series of

rectangular plates interconnected along longitudinal joints. The

analysis was developed using elastic plate theory for loads normal to

the plane of the plates and using two-dimensional plane stress theory

for loads in the plane of the plate. The analysis is limited to

straight prismatic box girders composed of isotropic plate elements

with no interior diaphragms and with simple end conditions.

Scordelis (10) and Johnston and-Mattock (11, 12) have utilized this

method in their studies of box girders. The inability of the analysis

to account for effects of interior diaphragms and anisotropic plate

elements such as transversely stiffened web plates prevents the appli-

cation of the above method to large size steel box girders. Recently

the method has been modified to take into account other than simple

support conditions as well as continuous box girders (13).

Since a box girder is made up of thin plates, the thin-

walled beam theories developed by Vlasov (14) and Dabrowski (15)

have been used as the basis of the refined methods of analysis.

Wright, Abdel-Samad, and Robinson (16, 17) extended Vlasov's theory

to consider stiffened plate elements as well as to include the effects

of interior diaphragms. Two methods were formulated. The "plate

element" method uses matrix analysis procedures. It treats the

structure as an assemblage of plate elements and utilizes a fourier

series solution. The "Generalized Coordinate" method formulates

equilibrium equations for the cross section and employs an initial

-3-
parameter solution method like that of Vlasov (14). This approach

permits consideration of flexible interior diaphragms and arbitrary

support conditions. A simpler version of this analysis to determine

distortional stresses has been developed based on an analogy to the

theory of beams on elastic foundations and is called the BEF

analogy (16, 17, 18).

Another analytical technique which has been applied to box

girders as well as many other structural problems is the finite element

method. A brief review of recent developments and the state of the art

on application of finite element method to box girders has b.een com-

pleted by Sisodiya and Ghali (19). The first application of the

finite element method to the analysis of box girders was by Abu-Gazaleh

and Scordelis (20) where they used six degrees of freedom at a nodal

point, three for plate bending (w, 8 , 8 ) and three for in-plane
X y
behavior (u, v, 8 ) • Multi-cell rectilinear box girders \vere solved
z
by Sawka and Cope (21) who represented the cells by in-plane elements

alone. William and Scordelis (22) developed a finite element program

to analyze box girder bridges of constant depth and arbitrary plan

geometry. Later Crisfield (23) developed a computer program for the

analysis of multi-cell, rectilinear or skew box girder bridges. His

analysis assumes symmetry about the middle horizontal plane_ of the

bridge.

In addition to these and other regular finite element analyses,

there are some modified versions. In the "Finite Segment Method"

(20) the basic structural elements used are formed by dividing each

web and flange into a finite number of transverse segmentso

-4-
Compatibility and equilibrium conditions are satisfied at selected

points along the four edges of each segment. The "Finite Strip

Method (24) is similar to the Finite Segment Method but with longi-

tudinal elements along the length of the girder.

Practically all the methods are confined to elastic analysis of

box girders. Very limited studies on the load carrying capacity or

ultimate strength of box girders have been made. Parr (25) in 1968

reported his work on the ultimate strength of box shapes having stocky

component plates and subjected to flexural loads only. In 1972

Corrado (26, 27) completed testing of two model box girders to failure

in bending and torsion, and formulated a method of estimating the

ultimate strength on the basis of research results on plate girders.

No mathematical or analytical procedure, however, was provided for the

evaluation of stresses at the component parts when some parts of the

box girder have been stressed beyond the elastic limit.

1.2 Objectives and Scope

The major objective of this work is to develop a procedure for

the evaluation of the ultimate strength of steel box girders. The

stress distribution and displacements of box girders in the elastic

and inelastic range of behavior are also sought.

The finite element method is chosen for this study because of

(1) the ability to include material nonlinearities for analysis in the

inelastic range, (2) its ability to incorporate diaphragm stiffness

for examining its effect, (3) its capability of handling different

-5-
isotropic and orthotropic components such as reinforced concrete decks

and stiffened plates if desired.

In this study, complete load displacement relationship is

investigated. To keep the extent of the study manageable in the non-

linear analysis, buckling of the component parts is excluded, as is

commonly done in stress evaluation of box girders. The method Hill

over estimate the ultimate strength if buckling occurs in early

stages of loading. Single cell, rectangular, prismatic and straight

box girders are the object of the study because there are experimental

results readily available on ultimate strength for comparison (26,

27, 28).

In the course of studying the ultimate strength, the behavior

and stresses of the box girders in the elastic range are also examined.

Results from tests and from other methods of solution are to be

compared.

-6-
 2. ELASTIC ANALYSIS OF BOX GIRDERS

2.1 Problem Formulation and Solution

The basic steps of the finite element theory and its appli-

cation can be found in many references (29, 30, 31). The finite

element analysis of an elastic continuum consist of (a) discretization

of the continuum into a mesh of finite elements; (b) evaluation of

the element properties; (c) assembly of the element properties into a

global stiffness matrix and incorporation of the boundary conditions;

(d) solution of the simultaneous equations. Only the necessary steps

of the analysis pertinent to this study will be presented.

2.1.1 Discretization of the Continuum

The basic structural elements used in this analysis of

straight box girders with rectangular cross section are rectangular

·· in shape. The elements are formed by dividing transversely and longi-

tudinally the \vebs and flanges as well as the diaphragms into an as-

semblage of small rectangular finite elements, Figo 2.1.

The selection of element shapes strongly influences the

simplicity of the problem formulation and solution. For box girders of

non-rectangular shape and with curvature, triangular, quadrilateral

and curved elements may need to be used. The subsequent evaluation

and transformation of element properties from the element coordinate

system into the global coordinate system require more work than that
-7-
for rectangular elements. Since the structure under study is rectan-

gular in shape and the primary objective is to obtain load displacement

relationship in the inelastic range of material behavior which would

require small sized elements, rectangular elements are chosen. This

way the element properties can be evaluated in the global coordinate

system with no need of transformation.

The size of the rectangular elements can be varied as desired

throughout the structure. In regions where the anticipated stress

gradient is high, such as locations near loading zone and supports,

a fine mesh of elements can be used. The thickness and material

properties of the elements can also be varied throughout the structure

to accommodate different plate thickness and various materials.

2.1.2 Evaluation of Element Properties

Element properties are expressed as the stiffness matrices of

the elements. There are mainly three methods for deriving the stiff-

ness matrix of a finite element: the displacement method, the equi-

librium method and the mixed method. In the first method, a dis-

placement field is assumed >vi thin the element and the element stiff-

ness matrix is derived from the minimum potential energy (29, 30).

For the equilibrium method, a stress field is assumed which satisfies

the equations of equilibrium and the element stiffness matrix is

derived from the principle of minimum complementary energy (30, 31,

32). The mixed method assumes both an equilibrium stress field and

a displacement field separately within each element, the element

stiffness matrix is derived from the variational principle (33, 34, 35).

-8-
 In the displacement method, if the assumed displacement field

is compatible, then the stiffness of the actual structure is always

overestimated and monotonic convergence to the correct solution from

below is ensured. Similarly, if the assumed stress field satisfies

the equilibrium of forces at the boundary, then monotonic convergence

to the correct solution occurs from above when the equilibrium method

is employed.

In the literature, the displacement method is used extensively

because it is relatively easy. The other two methods usually result

in a greater number of total degrees of freedom and greater semi-band

width of the stiffness matrix, thereby increasing the computational

effort (30, 35). It was therefore decided to use the displacement

method for this study, although in some cases the displacement method

gives slightly less accurate results of stresses when compared with

other methods.

The displacements [f} at any point within the element are

approximated by shape functions [N] associated with the generalized

coordinates [u} which are the nodal point displacements.

u.
1

·J u.
J
(2.1)

With the displacements known (or assumed) at all points within the

element the strain [e} at any point can be determined.

[e} = [B] [u} (2.2)

In the equation [B] is a matrix relating the nodal point displacements

to the element strains. It is obtained through differentiating the

-9-
shape functions [N] a~d rearranging them. It depends on the geometri-

cal dimensions of the finite element and is related to the type of

element and displacement field selected. Matrix [B] is independent of

material properties.

From the material constitutive law, the stresses at a point

are given by

(2. 3)

\{here [D] represents the elasticity matrix containing appropriate

material properties, (e } is the initial strain vector, (o } is the

0 0
initial stress vector, and (o} denotes the stresses within the element.

By applying to the element the virtual work principle or the

theorem of minimum potential energy, element stiffness matrix [k ] is

e
obtained as

[k
e
J ~ [B]T[D][B] d(volume) (2 .4)
volume

In evaluating the element stiffness matrices for this study,

it is assumed that the girder is made of thin walled members so that

Kirchoff's assumption is valid: that plane sections normal to the

middle surface of the plate remain plane after deformations. The

inplane and bending displacements are assumed to be small in com-

parison to the dimensions of the box girder. This implies that the

additional forces due to change of geometry are neglected.

Plate Bending

The plate bending behavior can be described by the out-of-

plane displacement, w, of the middle plane. Other parameters to

-10-
ensure at least an approximate satisfaction of slope continuity are

rotations about the x-axis (8) and rotations about they-axis (8 ).

. X y
By using the sign convention shown in Fig. (2.lb), the displacement

field which describes the bending deformations can be expressed in

vector form as

w w

[f} = 8X _(ow) (2.5)

oy

8 (ow)
y ox

Evaluation of the properties of elements which have the dis-

11
placement w11 as the nodal parameters are described in Refs. 29 to 31

and 36 to 38. Generally, higher order elements give improved accuracy

~;vhen fe\v elements are used. These elements may better satisfy the

boundary conditions and the displacement field may be closely approxi-

mated, but more time is usually required to generate the element

stiffness matrices.

In this analysis the ACM (Adini, Clough and Melosh) plate

bending element stiffness is used (29, 37)o The ACM element is non-

conforming in that slope continuity is not satisfied along the

boundaries except at the nodal pointso However as the number of

elements is increased the solution converges to the correct value.

Comparisons for plate bending elements are given in Refs. 36, 37 and 39.

In the ACM element a polynomial expression is used to define

the displacement field '\v" in terms of twelve parameters.

-11-
 w(x,y)

(2. 6)

the element stiffness matrix is obtained using Eq. 2.4 and it is

given explicitly in Refs. 29 and 37.

In-Ptane Behavior

The in-plane behavior of an element includes the displacements

u in the x-direction and v in tfie y-direction, the normal strains €

X

and € in these directions, and the shear strain y • The evaluation

y ~

of element properties for in-plane behavior are described in Refs.

29, 37 and 40.

In this analysis the "linear strain rectangle" element pre-

sented by Clough (42) is used for which the displacement polynomials

are

u(x,y)
(2. 7)

In the elastic range of material behavior, the elasticity

matrix for in.:.plane displacement is given by

1 \1 0

[De] E
\1 0 0 (2 .8)
2
1-\1
1-v
0 0
2

or for more general cases, the compliance matrix is given by Eq. 2.9.

-12-
 ·Dll D12 Dl3

[D] = D22 D23 (2;9)

(Sym)
D33
By using Eq. 2.4 and Eq. 2.9 ilie element stiffness matrix [ki] for in-

plane behavior is derived and is given in Table 2.1. A similar but

slightly different stiffness matrix has been presented in Ref. 42

resulting from the same displacement function. The difference is due

to the condition that the element D of Eq. 2o9, which is nonzero

13
in the nonlinear range of mate~ial behavior, was ommitted in Ref. 42.

This particular term is essential for the inelastic analysis in this

study (43).

Superposition of In-Plane and Plate Bending Behavior

With the assumption that the displacements are small, the in-

plane behavior and the out-of-plane behavior of an element are un-

coupled. Thus the total stiffness matrix of an element [k

e
J can be

obtained by direct combination of the in-plane stiffness matrix [k.J

~

and the plate bending stiffness matrix [kb]. In this analysis evalu-

ation is made first for the in-plane stiffness then for the plate

bending stiffness for the nodal points i, j, k, 1 sequentially (Fig.

2.lb) to obtain the total stiffness matrix for the element. Nodal

point displacements are in the order of

u u
v v
u. w w (2.10)
1

eX -~
oY
ey 00
ox
i i
-13-
2.1.3 Assembly and Solution

The individual element stiffness matrixes [k ] are trans-

e
formed from their local coordinate system into a global coordinate

system [k J
e g

[T]T [k
e
J [T] (2 .11)

where [T] is the transformation matrix for nodal point forces from

local coordinates into global coordinates (29).

The total global stiffness matrix [K] of bhe box girder is

obtained by summation.

[K]
e J
vol
[T]T[B]T[D][B][T] d(vol) (2.12)

The total stiffness matrix relates the forces [P} at the nodal points

of the structure to the displacements [6} of the nodal points (30).

[KJ[6} = [P} (2 .13)

In this analysis the parallel nature of the local and global

coordinate system is utilized (Fig. 2.1) . Element stiffness matrices

have been expressed directly in global coordinates rather than in

local coordinates. This procedure omits the step indicated by

Eq. (2.11).

From Eq. (2.10) each nodal point of an element will have five

degrees of freedomo For the nodal point at the junction of two per-

pendicular planes such as the junction of a web and a flange, it will

have six degrees of freedom. By adopting a proper sequence for

numbering the nodal points such that the maximum difference in the

nodal point numbers of the finite elements is minimized, advantage

-14-
may be taken of the geometry of the structure to yield the minimum

band width possible for the total stiffness matrix [K]. In the com-

puter program (44), advantage is taken of the prismatic nature of the

structure by numbering only the nodal points on the cross-section

X~ 0. Nodal point numbers for all other specified sections along the

length are automatically generated through computer programming.

The generalized nodal point forces {P} include initial

stresses and body forces. These forces can be applied through nodal

points at any place on the structure. Distributed loading can be

allocated to the nodal points using the consistent load vector

concept (29).

Boundary conditions are handled through constraining the cor-

responding degrees of freedom. Any support condition can be closely

approximated through this property of the finite element method. If

other methods such as thin walled elastic beam theory or the analogy

based on beam on elastic foundation are used, supports are assumed at

the neutral axis of the girder. In a short span girder the effects

of the support condition will be pronounced.

The Cholesky decomposition and backward substitution method

is used to solve the large system of equations. This is the most

time-consuming part of the problem in the computer.

2.2 Evaluation of Formulation

2.2.1 ~vo Degrees of Freedom versus Five Degrees of Freedom

As has been pointed out earlier each element nodal point has

five degrees of freedom. The results of the finite element analysis

-15-
are the in-plane and out-of-plane displacements of these nodal points

and the corresponding in-plane and plate-bending stresses of box

girder component plates. In this analysis loads are assumed to be

applied either through the diaphragms or at the junctions of the webs

and flanges, and the out-of-plane displacements are considered re-

latively small. This implies that, for the structures and the

loading conditions under study, a two degree of freedom condition

instead of the five degree of freedom assumption may be employed.

For a two degree of freedom case the band width of the total stiff-

ness matrix, a very important factor in the solution time, is only

about 1/3 of that for the five degree of freedom case with the same

mesh divisions. Thus the solving of problems would be much more ef-

ficient using only two degrees of freedom if its use can be justified.

To investigate this, five problems are solved by using both

two degrees and five degrees of freedom at a nodal point. Problem 1

to 4 use the same specimen, a single span rectangular composite box

girder (Girder Dl of Ref. 28) with concentrated loads applied at the

mid-span, Fig. 2.2. Problem 5 is a 100-inch long simply supported

wide flange beam of W8 x 31 cross-section, Fig. 2.3. The details of

the problems are summarized in Table 2.2.

·-
In Table 2.2 the results are compared for maximum deflec-

tions, maximum stresses at the top and bottom flanges and the central

processing time required by the CDC 6400 computer to solve the

resulting simultaneous equations. It is obvious that, for these t\vo

structures, the two degree of freedom and five degree of freedom

elements give practically identical results of stresses and

-16-
deflections. The maximum difference for all these example problems is

less than one percento The computational effort of the computer (CP

time) for the five degree of freedom cases, however, is ten to fifteen

times higher than that for the two degree of freedom system cases.

The number of operations in solving the simultaneous equations

2
resulting from Eq. (2.13) is proportional to NB where N is the total

degree of freedom and B is the semi-band width of the total st.iffness

matrix. For a given structure and finite element discretization, the

total number of degrees of freedom and the semi-band width both are

approximately 2.5 times higher for the five degree of freedom system.

This indicates that the solution time of the computer is about 15

times more for the former case. TI1e number of computer input data

cards, on the other hand, is exactly the same for both cases.

That the results from the two and five degree of freedom

systems are practically identical for the structures of Figso 2.2 and

2.3 is further indicated in Figs. 2.4 and 2~5. Figure 2o4 shows the

deflection profile of the bottom flange to web junction of the box

girder under bending and under torsion. Computed values have to be

superimposed on each other in the figure along the entire half length

of the box girder. In Figo 2.5 the results from the two and five de-

gree of freedom cases again fall on top of each other. Also shown in

the figure is the stress distribution pattern computed by the simple

beam (MY/I) theory in which the effect of shear on the stresses can-

not be included. The influence of shear lag effect in the flanges is

revealed by the results of the finite element analysis. This will be

discussed later.

-17-
 For both the box girder and the wide flange beam of Figs. 2.2

and 2.3 where the loads are applied at the edge of the plates, the

plate bending stresses from the five degree of freedom system are

small. This justifies the use of two degree of freedom formulation

in this analysis. This condition will be demonstrated further in

later examples.

2.2.2 Mesh Division Along the Length

Rectangular elements which have been widely used in two

dimensional elasticity problems give mediocre results in beam analyses.

when the ratio of the element length to its width (a/b) is greater

than unity (29). The deterioration in accuracy becomes more drastic

when the aspect ratio (a/b) increases from 1: 1 to 4: 1. This condition

necessitates a finer mesh division along the length of the beam when

finer mesh divisions are adopted for the cross-section.

The effect of the element aspect ratio on the accuracy of

results is not as marked for box sections as for beams with rectangular

cross-sections. This is because the longitudinal forces are mainly

resisted by the top and bottom flanges where the rectangular elements

replace the continuum better than the elements in the web do. To

demonstrate this effect of the element aspect ratio, simply supported

beams with three different cross-sectional shapes are analyzed. The

cross-sections are a rectangle, a wide flange and a box as are shown

in Fig. 2.6. The length of the beams is such that the resulting

element aspect ratios vary from 1:1 to 4:1.

The results are summarized in Table 2.3, where the deflections

and stresses are compared with known values from beam theory, including
-18-
the effects of shearo The accuracy of using rectangular elements is

poor for the rectangular beam even when the aspect ratio of 1:1 is

used. For the w·ide flange and the box shapes, the accuracy for 1: 1

elements is quite good, and for 2:1 elements it is acceptable, con-

sidering that a very coarse mesh is used for the analysis.

In this example of beam bending without torsion, the wide

flange and the box shape behave identically according to the beam

theory. The differences in deflections and stresses are due to the

fact that the box has double the web area of the wide flange beam.

For steel box girders, the webs are usually slender and small in area

relative to the flanges. The results of the finite element analysis

using rectangular elements of a moderate aspect ratio can be expected

to be better than those of Table 2.3.

To explore the effects of mesh division along the length

and the element aspect ratios further, the box girder of Figs. 2.2a

and 2.2b is analyzed using a different number of mesh.divisions along

its lengtho Both the simple bending and the pure torsional cases are

investigated. The stresses and deflections at some points are plotted

in exaggerated scale in Figs. 2,7 and 2.8 against the number of mesh

divisions along the length of the beam. Also shown are the aspect

ratio of the rectangular elements for the web. For all the stresses

and deflections the computed values are "stabilized" when only a very

few divisions along the half-length are used. For example, the

maximum deflection under bending for 10 kip applied load, obtained by

nine divisions along the half-length, is 0.04094 inch compared to

0.04123 inch by 30 division along the half-length. The difference is

-19-
only 0.67%. The compqtational effort, on the other hand, is more than

five times higher in the latter case.

Therefore, in evaluating the behavior of box girders it is

only necessary to have a moderate number of divisions along the length

of the structure.

2.2.3 Mesh Division Across the Webs and Flanges

One of the advantages of the finite element method is that

the shear lag effect is automatically included in the analysis. This

has been indicated earlier, and further examination will be made later.

Obviously the finer the element sizes across the cross-?ection, the

more accurate the stress and deflection values are obtained.

How fine the elements should be is evaluated using different

mesh divisions for the cross section of the box girder of Fig. 2.2

under simple bending and pure torsion. Ten divisions along the half-

length is arbitrarily chosen. First the web is divided into three

equal elements \vhile the number of divisions across the flanges be-

tween the \vebs varies from one to eight. The results are tabulated

in Table 2.4. For all simple bending cases, the computed deflections

remain almost the same, changing from 0.03829 inch per 10 kip of load

for one element across the flange to 0.03876 inch for the same load

for eight elements across the flange. The computed flange stresses

improves around 2%. The computational time of the computer, on the

other hand, has increased 15 times. A similar situation exists for

the pure torsional loading cases. The change in deflection estimate

is very minor \vhile the stress values improve 15 to 25%.

-20-
 The results of increasing mesh divisions across the web

are also similar. Table 2.5 lists the deflections and stresses of

the same box girder with two flange elements between the web and one

to six web divisions. As the number of divisions increases, better

results are obtained. The change (improvement) of results is at a

higher rate than that for Table 2.4 where the number of flange

mesh divisions is increased.

Usually mesh divisions across the webs and across the flanges

are chosen such that the resulting elements have_a moderate aspect

ratio (a/b). Therefore, fine elements in the web accompany fine

elements in the flange. The computational effort for a fine mesh

system often is fifteen to twenty times more than for coarse mesh

divisions, while the accuracy of stress values increase by five to

ten percent. Figure 2.9 depicts the relative increase of accuracy

and computational time for the box girder under study. An approximate

mesh division must be deduced from the importance of accuracy and

availability of computer capacity and time. For the box girders under

study, a flange mesh division of four with a web mesh division of

three to four appears to give fairly good results without consuming

too much computer time.

-21-
2.3 Comparison of Results with Existing Solutions

In order to check the validity of the assumptions and the

accuracy of the analysis, results of a multicell box girder, a simple

wide flange beam, a thin flange deck and a composite box girder are

compared with the results obtained by other investigators.

2.3.1 Multicell Box Girder

A two span continuous bridge with a rectangular cross section

of three cells is analyzed. The bridge has been analyzed by Scordelis

using folded plate theory, the rinite segment method and the finite

element method (20). The dimensions and loading conditions are shown

in Figs. 2.10a and 2.10b. The box girder is symmetric about the middle

support thus a 60 ft. span cantilever with a simple support at the

free end represents half of the bridge, Fig. 2.10c. nvo different

mesh divisions across the section have been used by Scordelis in his

finite element analysis and are also adopted here, Fig. 2.10d. Along

the length, eight divisions are used for a quarter of the bridge,

Fig. 2.10c, whereas only seven divisions at different spacing have been

employed in Ref. 20. Five degrees of freedom for each nodal point

are used in this analysis.

The resulting vertical deflections along the top of the loaded

web are shown in Fig. 2.11, together with the results of the folded

plate theory and the two finite element models from Ref. 20. The

vertical deflections of the bottom flange at cross sections 17.5 ft.

and 30 ft. from the interior support are given in Fig. 2.12. From

Fig. 2.11 it can be seen that the deflections along the loaded web

-22-
agree fairly well for all the methods considered. The results from

this study are slightly higher, with a maximum difference of less than

six percent of the folded plate theory values. The relative vertical

displacement of the flange at the exterior webs are also slightly

higher from this analysis, Fig. 2.12, but again the agreement among

the results of all the methods is quite good.

There are a number of factors which contributed to the dif-

ferences between the results of this analysis and those from the

finite element method of Ref. 20. First the polynomials selected in

the reference for the displacement field for out-of-plane deflection

and for in-plane behavior are different from Eqs. (2.6) and (2.7),

respectively. For in-plane behavior in addition to u and v, a third

term for average rotation about the z axis (8 = l [( 0x) c2Y ]) is

zi 2 ox J_. oY·
J_

considered in Scordelis' analysis. The six degree of freedom per

nodal point system describes the deflections better than the five

degree of freedom system used in this analysis. Secondly, the mesh

divisions along the length are not the same. Nor are the loading

conditions. In this study, the concentrated load is applied at a nodal

point whereas in the reference the load is spread over two fine

divisions of 6 inches each straddling the load point. Third, the

support and diaphragm conditions are different. Complete fixity at

the interior support and actual diaphragm thickness with a support

restraint against vertical displacement of the bottom flange are the

conditions of this analysis. The conditions that .. are used are not

given in Ref. 20 for exact evaluation.

-23-
 It must be pointed out that two degrees of freedom formulation

of finite element analysis does not provide accurate results for this

relatively thick walled concrete box girder. This is mainly due to

the high plate bending rigidity of the plate elements which cannot

be considered in a t~vo degree of freedom formulation. In evaluating

the plate bending rigidity of finite element, the elements in the

compliance matrix contain terms to the third power of the plate thick-

ness. Therefore only ~vhen the plate thickness is small, or when there

is little out-of-plane bending, should the two degree of freedom

system be used. To show how the thickness of a box girder effects the

selection of two degrees of freedom and five degrees of freedom the

same multicell box girder bridge of Fig. 2.10 is analyzed after it is

transformed into an equivalent steel box girder. The cross section

is transformed by using a ratio of elastic modulus n = Es/Ec = 10.

nvo loading conditions are examined. In the first one, two 500 kip

loads are applied symmetrically to the cross section to cause simple

bending. In the second case two antisymmetric 500 kip loads formed

pure torsional moment of 14,000 kip-ft. Mesh 1 of Fig. 2.10c is used

with both the two degree and the five degree of freedom system.

The resulting maximum deflections and stresses under the loading point

are compared in Table 2.6.

For the concrete box girder, the difference in deflection

between the results of the two systems is 13.8% for simple bending,

~vhich causes plate bending in the direction of the girder, and is 66.1%

for pure torsional loading which generate plate bending in both longi-

tudinal and transverse directions. When the cross section is

-24-
transformed into a steel section, the difference between the two and

five degree systems is only 0.46% and 2.43% for simple bending and

pure torsional loading respectively. For this reason, the multicell

concrete box girder bridge is analyzed using five degrees of freedom

per node, and the deflections are comparable to those obtained by

Scordelis as shown in Figs. 2.11 and 2.12.

Comparing normal stresses, using the box girder bridge of Fig.

2.10 again, the values obtained from this analysis agree well with

those obtained from Ref. 20. Figure 2.13 compares the normal stresses

at the edge of the flange along the top of the loaded web. The maxi~

mum difference is at the load point where the element sizes and the

manner of load application are different for the analyses considered

as has been pointed out earlier. Between the interior support (the

left end of Fig. 2.13) and the load point, the results are practically

the same. Bet\veen the end support and the load point, the computed

stresses are smaller than those of Ref. 20, definitely due to the

rigidity of the end diaphragms and also possibly due to the support

condition.

2.3.2 Wide Flange Beam

A simply supported wide flange beam (W8 x 31) of a 100-inch

span is analyzed. The same beam has been studied to compare two and

five degrees of freedom in Sect. 2.2.1. Due to symmetry with respect

to mid-span, to the longitudinal axis and to the web, Fig. 2.3, only

1/8 of the beam need be considered in the finite element analysis.

This implies that the load is applied vertically do\vmvard from the mid-

height of the web and the supports are at the longitudinal axis of
-25-
the beam. This load and support condition make the results more com-

parable w·ith the beam theory stresses and deflections. The Poisson's

ratio is taken as zero to make it comparable with the beam theory.

Because of the nature of the problem, two and five degree considerations

give identical results, Table 2.2, thus the former is chosen. Ten

divisions along the half span and six elements in the quarter of the

cross section (three in the flange, three in the web) are used, adding

up to 60 elements and 165 total degrees of freedom.

The computed deflections and the stresses along the centerline

of the bottom flange are plotted in Fig. 2.14 together with the results

from the beam theory. The beam theory deflections include the contri-

butions of shear which is 7.5% at the centerline. Although the element

aspect ratios are high (4:1) the deflections by the finite element

analysis compare satisfactorily with those from the beam theory. The

stresses from the two methods of analysis are practically identical.

When 40 mesh divisions along the half of the beam length is used, the

resulting deflections are practically the same. The maximum difference

is at the centerline of the beam and is less than 1% from the results

of Fig. 2.14.

2.3.3 Thin Flange Deck

One of the several tests performed by Schmidt (45) on plexi-

glas model flange decks is analyzed by the finite element method of

this study. The deck, Fig. 2. 15a is 0.405 em (0.160 in.) thick, has

a span of 157 em (61.8 in.), has t~vo inverted tees as webs and bottom

flanges, and is loaded by two concentrated forces directly above the

mid-span. Modulus of elasticity and the Poisson's ratio for the top
-26-
 2
flange deck and the web are 33,700 kg/em and 0.384, and are 33,500
2
kg/em and 0.376 for the bottom flange respectively.

Because of the double symmetry, only one-quarter of the·

structure need be analyzed by the finite element procedure.

Furthermore,because the bottom flange-to-web junction does not deflect

laterally, simplification can be made in discretization. Two different

mesh divisions across the cross section are used, being four and thir-

teen divisions of the top flange as shown in Figs. 2.15b and?.. 15c.

Along the length very fine meshes are adopted near the load and support

points and relatively coarse ones in between, adding up to 21 divisions.

The total number of elements, nodal point degrees of freedom, and cor-

responding semi- band w·idth of the stiffness matrix are all given in

Fig. 2. 15.

The results of two degree of freedom formulation are com-

pared with the test results and Schmidt's theoretical values. In

Fig. 2.16 the normal stress in the top flange 1.0 em (0.394 in.) away

from the loading point are shown, non-dimensionalized using a beam

theory bending stress at mid-span (45). The shear lag effect of the

thin flange deck is very pronounced. Both the finite element analysis

and Schmidt's procedure give good estimate of the normal stresses.

The finite element method results are 3 to 6% lower than the test

results. The displacements including the effects of shear are also

obtained by the finite element analysis, but no deflections are given

in Ref. 45 for comparisons.

To inspect the influence of plate bending rigidity, the

structure is analyzed using four mesh divisions across the top

-27-
flange and the five degree of freedom formulation. Since the flanges

and the webs are relatively slender (for example, the height to

thickness ratio of the web is 8.99/0.304 = 23.6) the results from

two and five degree of freedom analyses are expected to be very close

to each other. This is shown in Fig. 2.17 where the computed top

flange normal stresses are compared. The maximum difference is only

less than 1% at the cross-section, 1.0 em away from the load and at the

junction of the web and flange.

Also shown in the above figure are the results obtained from

the thirteen mesh divisions of the top flange compared with the four

mesh division stresses. The analysis provides almost the same results,

indicating that few divisions are required to capture the behavior

of the structure.

The example indicates that, for thin walled structures, a two

degree of freedom formulation with moderately fine mesh divisions will

provide fairly accurate results, including the effect of shear lag and

shear deformations.

2.3.4 Composite Box Girder

The finite element analysis is applied to a composite box

girder which is sho\vn in Fig. 2. 18. This box girder is specimen Dl

of Ref. 28 and has been taken as an example in Sect. 2.2 for the

evaluation of two or five degree of freedom analyses as well as for

determining the importance of mesh divisions. It is one of the main

structures to be analyzed later for the examination of inelastic behavior.

-28-
 The box girder has a span length of 10 feet and an overhang

of 2 feet at one end. It has one-sided transverse web stiffeners

spaced almost equally throughout the entire length. The 0.076 em

(5/64 in.) thick web has a slenderness ratio of 158. The bottom

flange thickness is 0.1875 em (3/16 in.) so as to prevent buckling

when loads are applied at the end of the overhanging part. Interior

plate diaphragms of 3/16 in. thickness are located at the supports, at

the mid-span and at the free end. Transverse loading stiffeners are

also added at these points to prevent local failure under load. The

concrete deck is 2.436 in. thick and is connected to the small steel

flanges of the web with very closely spaced shear connectors to ensure

complete interaction between the steel portion and the concrete deck.

The modulus of elasticity and Poisson's ratio are 29,600 ksi and 0.3

respectively for the steel portion. The average concrete modulus of

elasticity and Poisson's ratio are 3700 ksi and 0.15.

A two degree of freedom formulation is used in the analysis

since it has been sho~m in Sect. 2.2.1 that this method generates

alrnost the same results as the five degree of freedom system. Mesh

divisions for the analysis are determined from the results of the

Sect. 2.2.2 and 2.2.3. The web is divided into five divisions, the

half flange is divided into two equal parts, and fourteen divisions

are used for the half span length when the load is applied at the mid-

point of the main span. The overhanging portion is disregarded for

this loading condition and only half of the simple span needs to be

analyzed. Spacing of divisions is close (2.5 in.) near the load point

and the supports, and 5.0 in. between. When the load is applied at the

-29-
cantilever end, 26 di~isions are used. Support conditions are made

as similar to the testing conditions as possible. In simply supported

cases, the vertical displacements are constrained along the support

line, and horizontal movement normal to the plane of the web is pre-

vented from one point. For cantilever loading, displacement of the

bottom flange along the length of the girder is also prevented in

addition to the above constraints. Actual thicknesses of the dia-

phragms are used in the analysis. Due to symmetry only one-half of

the cross-section is considered. The loads are separated into pure

torsion and pure bending and the resulting stresses and deflections

are superimposed to give the loading condition. By using two degrees

of freedom per nodal point, the stresses in the elements are assumed

to be constant throughout the thickness.

In the analysis the contributions of the transverse and the

longitudinal stiffeners are not considered. Similarly the orthotropic

properties of the reinforced concrete deck are also neglected. All

these can be incorporated into the analysis for more detailed results

if desired. Effect of changing the modulus of elasticity of the

concrete deck is examined in Sect. 2.4.

In Fig. 2.19 deflections under the web along the bottom flange

are compared. Shmvn in the figures are the results from the finite

element analysis using the two degrees of freedom per node, the test

results, and values by the thin walled elastic beam theory, including

shear deformations and warping rigidity of the girder. Figure 2.19a

is for loading at the mid-span with an eccentricity of 7.688 inches to

the vertical line of symmetry. The methods of analysis predict the

-30-
deflections rather closely. The maximum deflection by the finite

element analysis is 2.7% less than the test results. Similar results

are obtained for other magnitudes of eccentricity and load in Ref. 28.

In Fig. 2.19b the deflection profile under the web is shown when the

load is applied at the cantilever end with an eccentricity of 4.188

inches. The results from the thin walled elastic beam theory are

lower, partly because rigid diaphragms are assumed in the analysis.

The test results are slightly higher in the main span than the computed

values by the finite element procedure and slightly lower in the can-
-
tilever. The overall agreement is deemed quite satisfactory.

TI1e normal stress distributions at a cross section (86.25 in.

from the left support) are shmvn in Fig. 2. 20a for the same loading

condition as for Fig. 2.19a. Some test results are available and

these generally agree with the results of finite element and the thin

walled elastic beam theory analysis. The same general agreement among

the two methods of analysis and test results is also evident in Fig.

2. 20b which shows the normal stresses due to a bending load only \vith-

out torsion.

In both Figs. 2.20a and 2.20b as well as many cross sections

inspected, the stresses from the finite elements method are lower than

the stresses by the thin ·walled elastic beam theory. Since these dif-

ferences occur not only under bending plus torsion but also under

simple bending, the cause can not be attributed to the influence of

warping torsion or torsional deformation. The mesh divisions of the

finite element analysis are sufficiently fine to avoid drastic inac-

curacy. The influence of t\vO or five degree of freedom formulation

-31-
is only very small for this structure, as it is indicated in Sect.

2.2.1. It is possible that the cumulative of these factors result

in the difference, but not probable. The main contributing factor

appears to be the basic assumption of girder depth and. "thin'' vall.

For all cases of analysis so far in this study, the depth of

a box girder or a beam has been taken as the centerline distance be-

tween the top and bottom flanges. This depth is larger than the

actual height of the web. Consequently, the computed displacements

and normal stresses are lower than those obtained by using the actual

web deptho For example, the maximum vertical deflection for simple

bending of the main span of the composite box girder is 0.003887 in.

per kip of concentrated load at the mid-span if the actual web depth

is used as the box girder depth. It is 0.004657 in. when the center-

to-center distance between flanges is adopted. This represents a 20%

difference with respect to the result from the actual web depth. The

corresponding maximum normal stresses at the bottom flange are re-

spectively 0.7710 and 0.6714 ksi having a difference of 13.5%. This

difference is reduced if the flange thickness is smaller, as for the

thin flange deck, Fig. 2.17, for which the difference amounts to 4%.

In all cases, the more realistic magnitude of deflections and stresses

lie between those two computed values and can be obtained only be

using a special web element around that region. Such an approach is

beyond the scope of this study. Reference can be made to some inves-

tigations on concrete bridges (37, 46, 47). Hereafter in this study

the center-to-center distance between the flanges is adopted so as to

be consistent with the thin Halled elastic beam theory.

-32-
2.4 Examples of Applications

The procedure of analysis heretofore developed in this study

aims at sufficiently accurate results and reasonably short duration of

conputer processing time. This capability serves as a basis for

analysis of straight box girders in the inelastic range of material

property. This also enables examination of box girder behavior in the

elastic range as it is influenced by girder geometry and loading con-

ditionso For example, the effects of diaphragm rigidity and diaphragm

spacing on the stress distribution and on box girder deflection and

deformation can be systematically examined under various combinations

of bending and torsion. A study of these effects is being conducted.

A number of simple examples are given below to illustrate the capa-

bility of the procedure.

2.4.1 Diaphragm Rigidity

From the thin-walled elastic beam theory and other procedures

of analysis, it is kno~vn that deflections and stresses are reduced if

a rigid diaphragm is provided at the load point~ How rigidity of the

diaphragm influences the maximum deflection and shear stress is de-

picted in Fig. 2.21. The box girder is that of Ref. 28 (Fig. 2.18)

with torsional load applied at the mid-span. All three diaphragms are

of the same thickness, varying from practically zero to 10.0 inches.

As the thickness is increased from zero, both the web shearing

stresses and particularly the maximum deflection reduce rapidly. Above

a certain thickness of diaphragm both the deflections and stresses are

not affected by further increase of thickness. For this box girder,

-33-
a 1/4 in. thick diaph~agm can be regarded as rigid with respect to

deflection.

The stresses in the diaphragm itself are influenced more by

the loading and supporting conditions than by the diaphragm plate

thickness. \vithin the elastic range of material property and without

plate buckling, the stresses are almost inversely proportional to the

plate thickness for a given loading and geometrical condition. Figure

2.22 shows the principal stresses at the 0.1875 in. thick diaphragm at

the loading point of the box girder of Fig. 2.18. Fine mesh divisions

may be used and lines of equal stress intensity can be plotted through

interpolation, but this is not done here.

2.4.2 Diaphragm Spacing

There are few existing guidelines for the determination of

daiphragm spacing. By using the same box girder of Fig. 2.18 both

with different number of equally spaced diaphragms, some results are

obtained from \vhich qualitative conclusions can readily be dra~;m. 1he

cases examined are: no diaphragm at all, diaphragms at the supports

only, and additional diaphragms at l/2, l/3, 1/4, l/5 and 1/6 of the

span lengths, respectively. All diaphragms are 0.1875 in. thick which

can be almost considered as rigid according to Fig. 2.21. Two dif-

ferent loading positions are investigated, one at mid-span the other

at quarter points.

The vertical deflections at the bottom of the web along half

span lengths are plotted in Fig. 2.23 for a typical torsional load

applied at the mid-span. The deflections are almost the same ~;.;rhen there

-34-
is no diaphragm at all, or when only support diaphragms are provided.

These deflections are many times of these when intermediate diaphragms

are used. Figure 2.24 shows an exaggerated scale of the deflection

for the cases with intermediate diaphragms. When the applied torsion

is bet~veen tv!O diaphragms (case L/ 3 and L/ 5) the deflections and con-

sequently rotations and distortion of cross-sectional shapes are

relatively large only between these two diaphragms. If a load is

applied at a diaphragm (cases L/2, L/4 and L/6) the deflections are

practically the same regardless of diaphragm spacing.

This phenomenon is further illustrated by Figs. 2.25 and 2.26

which are the deflection profiles of the box girder under torsional

loads at the quarter points. When there is no diaphragm between

supports, the deflections are many times higher than those when inter-

mediate diaphragms exist, Fig. 2.25. Whenever the load is between two

intermediate diaphragms (case L/2, L/3, L/5 and L/6, Fig. 2.26) the

deflections are relatively large only between the two adjacent

diaphragms. Beyond these adjacent diaphragms, the deflections (and

rotations and distortions) are practically the same as those when

there is a diaphragm at the load point (case L/4). Obviously, the

closer the diaphragms, the smaller the deflection between two diaphragms

~vhen load is applied therein.

The effect of diaphragm spacing on the stresses in the box

girder follow the same pattern of deflections.

It must be pointed out that diaphragm spacing has little

effect on bending of box girders. For example, the maximum deflection

-35-
of the box girder for a unit bending load at mid-span changes from

0. 00386 in. >vith diaphragms at the ends to 0. 00381 in. >vith diaphragms

at one-sixth points of the span. The corresponding normal stresses

at the bottom flange are 0.571 and 0.570 ksi, respectively.

2.4.3 Web Slenderness and Concrete Flange Rigidity

Further applications of the analysis are made to investigate

the effects of component dimensions or rigidity on the deflection

and stresses of the composite box girder which is being studied. The

web thickness is taken as the v~riable, changing from 0.060 in. to

0.240 in. and making the web slenderness ratio 200 and SO, respectively.

A simple bending load case and a pure torsional load case are investi-

gated separately. Loads are applied at the mid-span of the simply

supported box girder. Two divisions of half the flange between webs

and three divisions of the web are selected with ten divisions along

the half length of the girder for the analysis. The results of maximum

deflection and stresses at some points at the mid-span are tabulated in

Table 2.7. For bending load~ as the web thickness gets thinner and

the web slenderness ratio increases from 50 to 200, the maximum de-

flection increases almost linearly. Both the top and bottom flange

normal stresses increase, with the lower flange increasing more be-

cause the neutral axis shifts up for more slender webs. The web

shearing stresses increase drastically for thinner webs, but the shear

flo\v only changes moderately. Under torsional loading, the behavior

is similar. The shearing stresses in the web increase fast \vith

increasing slenderness ratio.

-36-
 In the analysis of the composite box girder in all the

examples so far, the modulus of elasticity of the concrete is taken

as 3700 ksi. The effect of modulus of elasticity of the reinforced

concrete flange deck is investigated by changing its value from 2500

ksi to 3000 ksi. The same mesh divisions and loading cases in the web

slenderness examination are used. The maximum bottom flange deflection

at the mid-span under a unit bending load decrease from 0.0420 in. to

0.0368 in. when the modulus of elasticity is increased from 2500 to

5000 ksi. This amounts to a 13% decrease of deflection for an

increase of 100% in Young's modulus. The normal stresses in the top

and bottom flanges increases from -0.273 ksi to -0.287 ksi and

decrease from 5.811 to 5.688 ksi, a 5% and -2% change. Under pure

torsional loading the percentage of changes are smaller. The

deflection decreases 7%, the web shear decreases 1%, and the top flange

shearing stresses decrease less than 1% for an increase from 2500

to 5000 ksi in.the modulus of elasticity. These results indicate that

the modulus of elasticity of the concrete is not a very important

parameter in the analysis of composite box girders especially in

torsion.

-37-
 3. ULTIMATE STRENGTH

3.1 Nonlinear Behavior

The ultimate strength of box girders is associated \vith non-

linear behavior of the box girders. The nonlinearity is either caused

by the nonlinear properties of the material or the nonlinear change

of the box girder geometry when subjected to load.

When the deflections of the box girder, particularly the de-

flections of the component parts of the box girder, are large enough

to significantly change the geometry of the structure, the equation

of equilibrium must be formulated for the deformed configuration. The

strains in the displacement equations may include higher order terms

for the large deflection analysis. The instability of the structure or

its components can be formulated as an eigenvalue problem. Numerous

studies have been carried out to investigate the geometrically non-

linear behavior and stress distribution of the box girder flanges and

webs (4, 5, 48, 49).

Material nonlinearity of box girders is due to nonlinearly

elastic and plastic or viscoelastic characteristics of the structural

material. Infinitesimal linear strain-displacement approximations

usually form an adequate basis for the evaluation of the displacements

and stresses under the condition of material nonlinearity. Ho\vever,

the complexity arises from plastic behavior under multiaxial

-38-
stress states. This fact together with the history dependence of

strains at different points of a structure makes even the solution of

a simple problem a formidable task. Consequently, existing solutions

to flange and web behavior due to material nonlinearity involve gross

simplifications. On the other hand, it has been shown that the incre-

mental step procedure for nonlinear displacement analysis is ideally

suited for structures with inelastic material properties (50, 51).

To evaluate the ultimate strength of box girders, both the

geometric and material nonlinearity should be considered. For case

studies \vhich analyze the behavior of a specific box girder from the

linear elastic range to failure, such thorough and complex examina-

tions may be carried out. For more practical estimates of nonlinear

behavior and load carrying capacity of steel box girders as a basis

of design, judgment must be made \vith regard to simplifications.

On the assumption that the instability of box girder components are

prevented and that the deflections are not excessive, attention in

this study is directed to the investigation of the nonlinear behavior

of box girders caused by the nonlinear properties of the material

alone.

Rectangular box girders are the objects of the study. Loads

on the girders are applied at junctions of component plates. The

finite element method is used for obtaining complete load-displacement

relationship considering material nonlinearity and by employing the

incremental approach.

-39-
 The essentials of finite element analysis in this study have

been presented in Chapter 2. It has been pointed out that for small

out-of-plane bending of component plates, the formulation of the

stiffness matrix by using two degrees of freedom per nodal point

provides sufficiently accurate results with only a moderate amount of

computational effort. Two degree of freedom formulation is adopted

for the nonlinear analysis.

3.2 Material Properties

In the elastic analysis, Chapter 2, the relationship between

stress and strain has been considered linear. The elasticity matrix

[D] of Eq. 2.3, thus can be readily evaluated. For the evaluation of

the nonlinear behavior of box girders due to nonlinear characteristics

of the material, these characteristics must first be defined. In this

study of steel and composite box girders, since normal stresses trans-

verse to the plane of the girder component plates are neglected, only

plane stress constitutive rules need to be described.

3.2.1 Stress-Strain Relationship

A linear elastic, limited plastic flow, and linear strain

hardening stress-strain relationship in uniaxial tension and com-

pression is assumed for the steel, Fig. 3.la. In the linear elastic

range, the steel elements of the finite element mesh obey Hooke's

Law (E) and Poisson's ratio(~). The yield stress level (cr ) and ilie
y
strain hardening modulus (E ) are determined by actual uniaxial
st
tension testing. In the analysis, any stress-strain curve different

-40-
from the one defined ~ere can be accommodated through either three

linear lines or the adoption of the Ramberg-Osgood equation.

Reinforced concrete is a heterogeneous material, but it is

cons ide red homogeneous in a macroscopi1c sense. Furthermore, it is

assumed to be isotropic in this analysis, although its orthotropic

nature can be handled in the finite element analysis. The uniaxial

tension-compression stress-strain properties are shown in Fig. 3.lb.

Under tension and compression, the stress-strain relationship is

assumed linear w·ith a slope of E • This assumption enables the uti-

. c
lization of the results of linear elastic analyses from Chapter 2

as the basis for nonlinear evaluation.

The concrete is assumed to crack at a tension stress 15% of

I I
the ultimate compressive stress, fc ; ft = 0.15 f
c
Under corilpres-

sion the proportional limit is taken as 0.70 f Between the pro-

c
portional limit and f at a strain of 0.002 in. per in., the stress-
c
strain curve is defined by a Ramberg-Osgood equation (46)

(3 .1)

where a and 8 are the corresponding stress and strain on the curve,

E is the modulus of elasticity of the linear portion, and m and n are

c
constants (46). Beyond 8 = 0.002, to the crushing strain of
Ef = 0.005, a horizontal straight line (zero slope) is adopted. By
I
Ref. 46, the values of the constants are taken as m = f c /0.002 Ec )
and n = 9. Any of these values can be adjusted atcording to actual

results of testing.

-41-
3.2.2 Yield Criteria

The elements are subjected to biaxial states of stresses, crx'

a and 'f The initial yield criterion for plane stress problems is
y xy
represented by a relationship of the form

F(a .. ) = K (3. 2)
l.J

where "F" is generally referred to as the loading function and K is

the yield condition, a known predetermined material constant. a ..

l.J
represent the stress tensor. Equation 3.2 represents a closed,

convex curve in a two-dimensional space (52). It is the initial

yield curve for steel or the failure envelope for concrete.

In this analysis Von-Mises yield criterion (52) is used for

steel. This criterion can be expressed as

2'
F(a .. ) a . cr + 3'f cr (3.3)
l.J x y xy o

where a is the uniaxial yield strength, or in terms of principal

0

stresses

(3.4)

An effective stress ae may be introduced and is defined as

2
+ cr 2 + 3'f 2
cr
e
==
1J cr x y xy
cr a
X y
(3.5)

By combining Equations ~3 and 3.5, Equation 3.3 can be written simply

as

a - a == o (3. 6)
e o

-42-
Initial yielding occurs when the effective stress.0 equals the uni-
. e
axial yield stress of the material, 0 . The von-Mises initial yield
0

curve is plotted in Fig. 3.2a.

The failure envelope of concrete under biaxial stresses has

been investigated by numerous researchers (43, 54, 54). Essentially

it can be simplified as in Fig. 3.2b. Failure can be either by

cracking (portion ABC of Fig. 3.2b) or crushing (portion ADC). Experi-

ments show that the strength of concrete under biaxial tension is al-

most independent of the principal stress ratio 0 /0 and is equal to

1 2
I
the uniaxial tensile strength, ft . Under biaxial compression, for

simplicity in this study, the von-Mises yield criterion with 0

0
= fc
is employed to approximate the failure curve. For tension-compression
I
and f are adopted except
:::·::.·::::::n:t:a:g(:~l::·:~)l:n:c:~~g In this portion the concrete
c

does not crack and the crushing failure curve is extended into the

regions.

3.2.3 Subsequent Yielding and Flow Rule

In Equation 3.2, it is defined that whenever the function F

becomes equal to the constant K, yielding will begin. After stress has

reached yielding at a point, subsequent behavior will depend on the

strain hardening characteristics of the material as in the case of

uniaxial stress state. For a perfectly plastic material, the initial

yield curve remains fixed. For a material that strain hardens, the

yield curve must change for continued straining beyond the initial

yield. Equation 3.2 can be generalized to describe subsequent yield

curves .. After yielding has occurred, K takes on a new value depending

-43-
on the strain hardening characteristics of the material. There are

three different cases for a strain hardening material:

oF
dF = da .. >0
l.J
Loading (3.7a)
oaij

dF =
.9!_ dcr ..
l.J
= 0 Neutral Loading (3. 7b)
l.J
oa ..
oF
dF = dcr .. < 0 Unloading (3. 7c)
ocr .. l.J
1J

If the material is unloaded and loaded again, additional yielding will

not occur until the new value of K is reached"

In this study, a work hardening material will be assumed with

an isotropic strain hardening (52). The value of K for the subsequent

yielding in the incremental theory of plasticity can be calculated

from the amount of the plastic work or from equivalent plastic strains.

Then the yield function, F, can be derived as a function of the

equivalent plastic strain, €p' (52,55).

F (a .. ) = H (fp) (3.8)
l.J

or using the von-Mises yield criterion, Eq. 3.6,

a - H (e ) = 0 (3.9)
e p

In the analysis the experimental uniaxial stress-strain relation-

ship is used for the evaluation of the functional relationship.

So far the yield condition and the loading functions have

been defined. Because of the dependence of the plastic strains on the

loading path, it becomes necessary to compute the increments of

plastic strains throughout the loading history. In order to express

-44-
the increments of plastic strain on the basis of the present state of

stress and the stress increment, another assumption has to be made.

The associated flow rule of the von-Mises yield criterion

which is the r;Prandtl-Reuss Equation" is assumed to be valid for the

steel, as well as for the concrete after crushing in compression. This

principle can be expressed as

= dA. _Q[_ (3.10)

Ocr ••
~J

where [E}p is the plastic strain increment and dA. is a proportional

constant. This constant defined for the case of von-Mises yield

criterion is

dA. (3 .11)

where cr is the incremental effective stress and H' is the slope of

e
the effective stress versus strain curve.

Usually, the incremental stress [~} and the incremental strain

[~} are directly related by an elasticity matrix. In the elastic

range,

[D] [e} (3. 12)

The elasticity matrix [D] is constant and is given by Equation 2.8.

In the inelastic range, the compliance matrix is called [Dep]. The

definition of [Dep} is based on

(3. 13)

a derivation similar to that used by Yamada (56, 57) and is sho\vn

below.

-45-
 The total strain increment (€} can be written as the summation

of elastic strain increment (s}e and the plastic strain increment

(3.14)

The elastic strain increment has been defined previously

(3.15)

For an isotropic material if (a} represent a stress state of an element

just bef?re it undergoes the additional strain (€}, the deviatoric

I
stress state (a} corresponding to (a} is

a X
a
X - 31 (ax + cry)

I
1
[a }= a a (a + a ) (3.16)
y 3 X y
y

'fxy 'f
xy
I
For the von-Mises yield criterion, [a } oF
Then Equation 3.10
Cu ...
~J
can be written as

[€}P = (a'} d;\ (3. 17)

From Equations 3.14, 3.15 and 3.17

I

(E:J = [De] (cr} + [a } dP. (3.18)

or
-1 I

(cr} = [De](E:} - [De] [a } dA (3.19)

The effective stress ae can be expressed in terms of the deviatoric

stresses as

2 '2 '2 '2

a
e
=3 (a
X
+ ay + a
X cry + 'fxy ) (3.20)

-46-
 • I
(2·a +a )
y X
• I
+ 2 Txy T
xy
] (3.21)

From Equation 3.16

2 a
. +a
. aX
X y
.
y + aX
2 a a (3. 22)
y

T T
xy xy

which, when inserted into Equation 3.21 results in

I T
2 ae a e 3 [a } [cr} (3. 23)

Equation 3.11 can then be written as

9
dA. 2
(3.24)
4 a H
e
I T
By premultiplying Equation 3.19 by [a} and using the vector (s} to
I

represent [De] [a}, Equation 3.24 becomes

2
4 cre H T . I T
--9~- dA. (s} [d - [a }· [s} dA. (3.25)

Defining

(3.26)

then

dA. =1 s"~:
[s}T £~} (3.27)

By substituting this back into Equation 3.19

• e 1 T ·
[o} = [D ] - s~·: [S} [S} [e} (3.28)

-47-
 Define

Then [cr} = [Dep] [€} is as expressed in Eq. 3.13.

In cartesian coordinates, Eq. 3.28 can be expressed explicitly

as

dcr
X
I
1
E

- \)
2
s 2
1
g-;'(
(Sym.) l de:
X

'V E sl s2 E s2
dcr de:
y
1 - .2
\)
s~·:
1 - \)
2 S7: y

s 2
sl s3 s2 s3 E 3
dT xy .J s-t~ S7: 2(1 -\)) S.,'~
de:
xy

(3. 30)

where
I I
E
sl 2(crx + \) cr y )
1 - \)

E
s2 z<cry + crX )
1 - \)

S = ---=E-,..
2
3 1 - v I

xy

The definition of S7< requires the value H1 which is the slope of the

effective stress versus effective strain curve. This slope may be

obtained from the uniaxial stress-strain curve.

-48-
3.3 Solution Technique

3.3.1 Incremental Method

In the finite element solution of linear elastic problems

Chapter 2, the answer of displacements are obtained by solving the

simultaneous Equations 2.13. No trial and error or iteration is re-

quired since the stress-strain displacement relationship is linear.

If the stress-strain relationship is nonlinear, solutions can not be

obtained directly. Either an iterative method for certain classes of

problems (29, 58, 59, 60) or a~ incremental procedure (30, 61, 62),

or their combination i.e. an incremental iterative method (46, 62,

63) must be used.

In the iterative procedure, the total load is applied to the

structural system. The stiffness matrix is derived based on the

characteristics of one point on the stress-strain curve of the

material. Thus equilibrium is not necessarily satisfied and iter-

ations must be performed. This is done through successive adjustments

of the elastic constants (E and v) in the elasticity matrix [D] or

[De]. Such adjustments can be handled fairly easily for problems

of nonlinear elastic material. For materials exhibiting plasticity,

the compliance matrix [Dep], Eq. 3.29, is fully populated and the

adjustments of the material property constants are extremely difficult

if not impossible. Consequently, the iterative procedure can not be

used here.

The basis for the incremental method is applying small incre-

ments of load successively and considering the structural behavior

-49-
as linear \vithin each increment of load, Fig. 3.3. The total stress

and strain under a given total load are obtained by summation.

For each increment of load [.0. P.} the relationship bet\veen

~

this load increment and the·increments in displacements [.0. 6.} can be

~

expressed as

[K . ][ .0. 6 . }
~ ~
= ( .0. P.}
~
(3.31)

for nonlinearity due to material properties only (64, 65). [K. J is the
~

assembled total stiffness matrix for the load increment. If the

initial load and displacement are [P 0 } and [6.} respectively, the

~

total displacement and forces at any stage are given by

(P.} (P } + l: ( t. P.}
~ 0 i ~

(3. 32)
[6.} = [6 } + ~ [t. 6.}
~ 0 ~ ~

The procedure is the most general technique available for

solution of elasto-plastic problems and is used particularly with the

flow theory of plasticity. The incremental procedure is comparable

to the Euler method of solving differential equations of initial

value problems. Therefore, methods such as the Runge-Kutta procedure

(66) or other procedures for the acceleration of the solution of

initial value problems are also applicable.

3.3.2 The Tangent Stiffness Method

The stiffness matrix [K.] of Eq. 3.31 can be derived by using

~

the tangent modulus, the secant modulus, or the initial modulus for

the compliance matrix and the element stiffness matrix (29, 30, 46,

63). When the tangent modulus is used with the incremental procedure,

-so-
the method is called the incremental tangent stiffness method (57, 61,

62).

The method was initially suggested by Pope (61) for the

solution of elastic-plastic problems by finite elements. It was then

adopted by Marcal and King (62). Pope used an iterative procedure to

determine the size of a load increment which initiates yielding at

an additional element. Yamada (5~ developed a closed form solution

for the determination of the load increment. In this study, a procedure

similar to that by Yamada is used. The size of the load increment

[6 P.} at every stage is determined as a result of the analysis itself

1. .

such that the load increment is sufficient to cause yielding in a

specified number of elements.

The procedure used in this analysis is summarized as follm-1s:

1. Apply an initial load vector, analyze the system

elastically as described in Chapter 2, calculate

displacements and stresses, calculate effective

stress (cr ) for each element.

e
2. Scale all the elastic values by a factor in order

to induce first yielding at the element of maximum

equivalent stress. Any additional load \V'ill cause

inelastic behavior of the structure.

3. Calculate according to the stress-strain curve the

elastic-plastic compliance matrix [Dep] and hence-

forth the element stiffness matrix [k ] for the

p
post-yield elements.

-51-
 4. Assemble the global stiffness matrix [K.], apply
1.

a dummy unit load vector, solve for the unknown

displacements and evaluate the corresponding

stress data.

5. Calculate the scaling factor (r) for each element

corresponding to its state of stress (elastic or

inelastic) in order for the element to reach

yielding, cracking, or ultimate strength.

6. Select a scaling factor (r ) such that a desired

m

number of elements will be yielded at that

increment of load, scale the stresses and strains

with this factor and obtain the total stresses,

strains, displacements and loads.

7. Hhen the slope of the load deflection curve is less

than a specified value, or if the load has

reached a required value, the procedure is

completed. Othenvise, return to step 3, and

repeat.

This procedure requires more computational effort than

needed for a load increment in the initial stiffness approach, but

the tangent stiffness method permits a significant increase in the

size of the load increment as compared to other methods.

-52-
3.4 Illustrative Problems

3.4.1 Wide Flange Beam

The solution technique is first applied to a fix-ended beam

which is 14 feet long with a W8 X 40 cross section, and is loaded at

the third points. It is one of the beams examined by Knudsen, Yang,

Johnston and Beedle (67) for the analysis of inelastic behavior of

beams. The beam is suitable for testing the procedure of this study

because the effects of geometric nonlinearity are minimal until just

before failure of the beam. This problem is also analyzed in Ref. 46.

The actual dimensions of the beam are as follows: flange

width and thickness, 8.06 in. x 0.552 in,; web thickness, 0.370 in.

and overall depth, 8.32 in. The modulus of elasticity and Poisson's

ratio are assumed as 29,600 ksi and 0.3 respectively. The stress-

strain curve is from testing a standard tensile specimen and shows a

trilinear relationship. The yield point is 37.8 ksi. Onset of strain

hardening occurs at a strain of 0.017 in./in. Tensile strength is 52

ksi with an average strain hardening modulus of 630 ksi (67).

The loading and the mesh size for the finite element analysis

are shown in Fig. 3.4 .. Due to symmetry along the span, only half of

the beam length is considered. Fifteen divisions are used along the

half span. The loads are assumed to be applied at the center of the

cross section so as to take advantage of the double syn~etry of the

cross section to cut down the total number of degrees of freedom and

the band width in the analysis. The half flange is divided into two

elements and the half depth into five. The center to center distance

-53-
between the flanges is considered as the depth. For this condition

of mesh division, there are 272 simultaneous equations in Eq. 3.31 and

the semiband width is 22. Other mesh divisions have also been examined

resulting in different number of equations and bandwidths. The final

result of ultimate strength or load carrying capacity of the beam,

however, differs very littleo

The load-deflection curve for the midspan, as obtained from

this analysis, is sho~m in Fig. 3.5, together with the test data from

Ref. 67. The curve can be approximated by three straight lines. The

first linear part corresponds to the elastic behavior of the beam when

stresses, strains and deflections are all proportional. The second

straight line portion starts when yielding commences at the flange at

the supports. As higher loads are applied, yielding penetrates from

the flange into the web \vhile strain hardening starts at the extreme

fibers. When a very small load increment causes a very large increase

in deflection, yielding at the supports has practically reached the

neutral axis and the plastic hinges have formed. In the analysis, at

this stage, 57 of the 105 elements have plastified and 14 have reached

the tensile strength of the material. The beam has attained its

ultimate strength or load carrying capacity.

The test results agree quite \vell tvith the predicted values

from the finite element analysis. The test results are slightly

lower due to the existence of residual stresses in the test beam, to

the conditon that the end supports are not 100% fixed, and to the

inherent characteristic of the displacement method in overestimating

the stiffness of a structure thus giving higher predicted strength.

-54-
The residual stresses, if known, can be taken into consideration in

the analysis. Without considering residual stresses, the simple

plastic theory predicts that first yielding occurs at 71.6 kips with a

midspan deflection of 0.30 in. The corresponding values by this

analysis are 65.9 kips and 0.313 in. being closer to the test results.

The effects of shear are not included in the simple plastic theory

but are automatically considered in the finite element analysis. The

load carrying capacity is 115.8 kips, 114.1 kips and 107.0 kips by

testing, this study, and Ref. 67 respectively.

The most predominent factor for an accurate determination of

the load-deflection behavior of the beam is the size of the load

increment. Logically, if one element is allowed to yield at one

increment of load, the most accurate result will be obtained for the

mesh division chosen. This ~vill also enable tracing the sequence of

yielding of elements in the correct order. However, such a procedure

requires many increments of load to determine the ultimate strength,

a matter of computational effort. To speed up the computation, the

load increments can be determined such that a prescribed number of

elements will reach or exceed the yield stress during a load increment.

As the load carrying capacity is approached, this method may not be

practical since very large deflections accompany yielding of even one

additional element. A second method of increment is imposed in terms

of .the maximum and minimum percentage of the total load. Higher

percentages of load. increments can be allowed at the beginning of

yielding, and the percentages should be reduced as the rate of

deflection increases for an increment of load. For this illustrative

-55-
example, the maximum percentage changes from 3% at first yielding to

0.4% at forming of the plastic hinges •

. This illustrative problem demonstrates the workability and

accuracy of the procedure for the elasto-plastic analysis to estimate

the ultimate strength of structures with negligible geometric non-

linearity.

3.4.2 Steel Box Girders

There are only very limited results from ultimate strength

testing or analysis of box girders. The two small model specimens

reported in Refs. 26 and 27 have been designed to observe the

influence of web yielding on the load carrying capacity of the box

girders. The webs failed by tension field action in resisting high

shear which generated relatively large out-of-plane deflection of the

web plates. The effects of geometric nonlinearity were therefore

prominent, in addition to the effects of the material nonlinearity.

For lack of test data, one of these model box girders is analyzed

in this study for further evaluation of the procedure of analysis.

The dimensions of the model box girder are indicated in Fig.

3.6. The load is applied directly above one web at the mid-length,

symmetrical to the beam span. The yield stress of the top flange, the

webs, and the bottom flange are 32.5 ksi, 30.4 ksi, and 31.3 ksi,

respectively. The corresponding tensile strength are 47.4 ksi, 43.4

ksi, and 45.6 ksi. For all cpmponent plates, the onset of strain

hardening is assumed to start at a strain of 0.012 in./in. and to have

-56-
a strain hardening mod?lus of 500 ksi. The modulus of elasticity and

the Poisson's ratio are 29,600 ksi and 0.3 respectively.

In the finite elements analysis, the transverse stiffeners

are not considered. Furthermore, for symmetry, the intermediate

X-diaphragm at the left half is ignored. The half span length is

divised into eight parts, with closer mesh divisions (1 in.) at the

support and loading point and relatively coarse ones (2 in.) in between.

The flanges between the webs are each divided into two equal parts, and

the webs into three. The cross-shaped diaphragms are approximated by

plates having the same thickness as the webs and are connected to the

box girder only at the four corners of the box. In the testing of

the model, the midspan X-diaphragm did not appear to be sufficiently

strong and failed before the attainment of the girder ultimate load.

Consequently two conditions are considered in this analysis, one with

the midspan diaphragm and one without.

The resulting load-deflection curves and the test results

are sho1m in Fig. 3.7. As anticipated,. the case of no loading

diaphragm has higher deflections even under relatively low loads of

800 and 1000 lbs. The midspan X-diaphragm appears to be effective up

to at least 1200 lbs. At 1400 lbs, the first definite sign of tangent

diagonal web bulging appeared in panel 6 at the loaded side (26, 27).

The effects of geometric change started to influence the load-carrying

capability and the result of this analysis considering the midspan

diaphragm underestimates the deflection. At 1600 lbs. the X-diaphragm

has failure so that the lmver curve in Fig. 3. 7 is valid for compar-

ison with test results. Again, the measured deflection is higher than

-57-
predicted because of the large deflections of the web plates. Actual

failure occurred at 1800 lbs. due to local failure of the transverse

stiffener at the load point. The computed load-deflection curve has

an upward turn around this load magnitude because of the strain-

hardening effect. By taking into consideration the failure of the

X-diaphragm but not the tearing of the transverse stiffener, Corrado

predicted a ultimate strength of 1700 lbs. whereas the finite element

approach estimated 1894 lbs.

The influence of the size of load increment, as pointed out

earlier, is apparently depicted in Fig. 3.7 by the upper curves. For

the same mesh division and the same specified number of elements to

be yielded at each load increment, the coarse load increments result

in a higher load-deflection curve. If, for example, four elements are

permitted to yield after an increment of load~ P. from load P.,

1+ 1 1

~ P;_l-l is determined by a scale factor v

• m
P.
1
= P.1 + ~ P. , , such
1• 1

that the fourth element will just reach yielding. The other three

elements will have stresses higher than yielding. This condition of

stress will be corrected through the total stiffness matrix of the

structure, but only during the next increment of load. The resulting

load-deflection curve is thus higher than the actual one. Therefore,

when permissible, finer load increments should be used, and a small

percent of load increase should be specified.

The example indicates that the procedure can be employed for

ultimate strength evaluation of box girders with acceptable results,

even when geometric nonlinearity exerts strong influence on the be-

havior of the box girder. An analysis of the other small model box
-58-
girder of Refs. 26 and 27 gives similar results. In fact, the com-

puted stresses in the box girders agreed quite well with the measured

values. Figure 3.8 shows the load versus shearing stress relation-

ship of a point in the web of specimen M2 of Refs. 26 and 27. Before

bulging of the web, the computed and measured shearing stresses are

practically the same. Even after the web deflections start to in-

crease, the predicted stresses are still acceptable.

3.4.3 Composite Box Girders

Two composite box gir~ers, Dl and D2, have been subjected

to bending and torsion and have been tested to failure (28). The

behavior of girder Dl in the elastic range is examined in detail in

Sect. 2.3.4. Its ultimate strength and behavior in the inelastic

range are analyzed here using the procedure of this study. The

geometry of the box girder is shown in Fig. 2.20. The web slender-

ness ratio and the stiffener spacing have been designed such that

web buckling would not occur. Consequently the effect of large de-

flection of component plates are minimized and the assumption that

only material nonlinearity takes place is valid, at least for a

major part of the analysis.

Some of the material properties have been given in Sect.

2.3.4. The yield stress and the ultimate stress of the steel parts

are 31.0 ksi and 44.0 ksi, respectively. The stress-strain relation-

ship is almost trilinear, as obtained from a tension coupon testing.

The strain hardening begins at 0.014 in./in. and the average strain

hardening modulus is 500 ksi. For the reinforced concrete top flange

-59-
the stress-strain curv~ is that which has been discussed in Sect. 3.2.1,

Fig. 3.1b, with an ultimate compressive stress of f 5.5 ksi.

c

The discretization of this box girder into a mesh of finite

elements has been examined in Sect. 2.2.2 and 2.2.3 and the mesh

divisions adopted for the elastic analysis are given in Sect. 2.3.4.

The same mesh divisions are used here for the ultimate strength

analysis.

Continuing from the elastic analysis of Sect. 2.3.4, the

effective stresses G of all elements are computed for the theoretical

e
yield load at which the highest stressed element in the box girder

reaches yielding. An incremental load is then applied and the new

effective stresses under the total load are evaluated according to the

current values of the compliance matrix [Dep], the element stiffness

matrix [k ], and the global stiffness matrix [k.]. For any of the
e 1

elements to each yielding, a scaling factor y can be determined by

using the formula (50)

1+ r2 + 4G 2 (F G 2)
ei e
r = 2
2 Gel
(3.34)

2 2
r= crel - 2 G (6 cr e )
e - (6 G )
e

where Gel is the effective incremental stress due to the incremental

load and is computed by using Eq. 3.20. F is the yield stress of

y
the material. 6 G is the increment of the effective stress from G ,
e e
due to the load increment. If at this stage a total of m elements

are specified to reach yielding, then the scaling factcr, r ' for
m
-60-
the mth element is used to determine the magnitude of the load

This process is continued for all elements which are not

yielded. For yielded elements as stresses approach the tensile

strength of the material, the same procedure of scaling for each load

increment can be applied. It is only necessary to replace the yield

stress F in Eq. 3.34 by the tensile strength cr . The appropriate

y u 1t

value of strain hardening modulus must then be used. For concrete

elements, the same procedure can also be used with its yield curve

as defined by Fig. 3.2b.

The load deflection curve of box girder Dl for the vertical.

deflection at the mid-span and under the web is given in Fig. 3.9.

The test data are also sho~vn. The correlation between the test data

and the analytical results of this study is very good. Except for a

few test points around the bend of the curve, which are higher than

the actual static loads of the test (28), the computed curve

practically coincides with the test data. No residual stress \vas con-

sidered in the analysis, and the actual magnitudes of these residual

stresses are assumed to be small (28). The analysis indicates the

effect of strain hardening at the ultimate load, just as the test

results have shown. The unloading portion of the box girder behavior

can not be described by the analysis. The incremental procedure is

stopped at the ultimate load level when the incremental deflections

are excessive corresponding to small incremental load.

-61-
 Specimen D2 at Ref. 28 is also analyzed. This box girder has

a thinner web and a more liberal spacing of transverse stiffeners when

compared to specimen Dl. The webs of the girder developed prominent

tension fields and lateral deflections of the web plates were not

small prior to the attainment of the ultimate load. The results of

the analysis, shown as the load deflection curve in Fig. 3.10 deviate

only slightly from the test results in the elastic range of behavior.

The difference increases as the load approaches the highest value.

At the ultimate load, the predicted value is 3% higher than the test

data. This is regarded as quite acceptable considering that geometric

nonlinearity is not incorporated in the analysis.

As it has been pointed out earlier, mesh divisions influence

the results. Finer elements permit more accurate predictions. It has

also been mentioned that the size of load increment affects the

computed deflections and stresses. Smaller load increments give magni-

tudes closer to the measured ones. Box girder Dl is analyzed by using

a relatively coarse mesh division but small load increments. The webs

are each divided into three parts compared to five for Fig. 3.9.

The top and bottom flanges between the webs are divided into two equal

parts instead of four. Ten divisions along the half length of the

box girder are used in place of 14. The ultimate strength obtained

is the same as that shown in Fig. 3.8. This indicates that, for the

determination of the ultimate strength of box girders, a relatively

crude mesh division can be used so long as the load increments in the

process of analyzing are reasonably small.

-62-
 The inherent -condition of the displacement method is that the

predicted stresses may be less accurate than the deflections. To

examine this, the variation of the normal stress at a point of box

girder Dl is obtained and is plotted in Fig. 3.11 for comparison with

measured values. The point examined is at the bottom flange, 3.75 in.

away from mid-span. In Figo 3.11, the data points are converted from

the measured strains according to the stress-strain curve of the steel.

Above 60 kips of applied load, the computed stresses increase because

of strain hardening, and the values deviate from the measured stresses.

In the range of elastic behavior and the first portion of the plastic

flow, the computed stress agree very well with the test data. It is

to be noted that the load increments in the inelastic analysis are

quite small, as can be seen in this figure.

From the finite element analysis, the conditions of the box

girder at all loads can be evaluated. Figure 3.12 shows an example

which depicts the directions and magnitudes of the principal stresses

in the elements of one web panel of girder Dl. At the load magnitude

of 56.84 kips, the element adjacent to the load has very high corn-

pressive stress and the lower four elements are under high tensile

stress. All the elements next to the centerline of the girder have

yielded. In fact, a number of elements have reached yielding, as

is shown in Fig. 3.13. The spread of yielding can easily be traced

throughout the entire loading sequence.

The significance of these illustrations on principal stresses

and yield spreading is that, when the residual stresses are not

-63-
negligible and their magnitudes are kno•~, their effects on the be-

havior of the box girder can be evaluated throughout the entire range

of the box girder behavior. Coupled with the predicted deflections,

these provide a reliable basis for design.

-64-
 4. SUMMARY AND CONCLUSIONS

The ultimate strength and the load-deflection relationship

of straight, rectangular box girders have been investigated. In-

stability of the box girder or its components is assumed to have been

prevented. The loads on the girder are applied either through

diaphragms or at junctions of webs and flanges. The finite element

method is used with an automatic mesh generation program; and the

displacement method is employed with the incremental load process.

In the investigation the elastic behavior of the box girders

is first evaluated and compared with existing solutions so as to

examine the procedure developed in this study. The inelastic analysis

is carried out based on the results of this examination. A very brief

example is given to show the capability of the procedure in developing

aids for design and analysis.

Conclusions which can be drawn from this study are the

following:

1. ~vo degrees of freedom per nodal point in the finite

element formulation of box girder analysis can give accurate results

compared to five degree of freedom formulation, if the out-of-plane

plate bending is insignificant. For the same finite element discre-

tization, the two degree of freedom system requires much less computer

processing time.

-65-
 2. For the box girders, fairly coarse mesh divisions

provide fairly good results for deflection. For a better estimate of

stresses, moderately fine mesh divisions are sufficient. Elements

should have a low· ratio of length to width, preferably not more than

2:1, depending on the fineness of the mesh division.

3. The finite element procedure incorporates the effects

of shear lag, shear deformation, diaphragm rigidity and spacing,

and boundary conditions. These can be used for the analysis of thin

flange deck as well as box girders.

4. The procedure provides an efficient and accurate method

for the evaluation of box girder deflections and stresses.

5. For composite box girders which have a concrete top

flange, the selection of the box girder depth would affect the results

of the analysis. Either the actual web depth or the centerline dis-

tance between the flanges may be considered as the box gi_:rder depth.

Test results indicate a depth in between the two values.

6. 1he rigidity of diaphragm affects the behavior of box

girders under torsion. The required thickness or rigidity of a

diaphragm to ensure maintaining of cross-se.ctional shape can be

easily estimated.

7. When torsional loads are applied between two diaphragms,

the deflection and rotation of a box girder are relatively large only

bet\veen these diaphragms. If torsional load is applied at a

diaphragm, the deflection and rotation of the box girder are

practically the same regardless of the diaphragm spacing.

-66-
 8. Diaphrag~ spacing has very little effect on the behavior

of box girders under flexural loading alone.

9. The load-deflection relationship of box girders can be

obtained accurately by the procedure of this study if plate deflections

of the web and flanges are not large. Girder deflections and cor-

responding stresses and strains can be determined for loads which

only cause elastic responses as well as for loads which would cause

large deflection of the girder and failure.

10. The ultimate strength or load carrying capacity of box

girders can be predicted from the geometry and material properties

of the box girder.

11. The estimated ultimate strength is higher than the

actual load carrying capacity. The amount of overestimation depends

on the finite element mesh division and the sizes of the incremental

load. Finer mesh divisions and more important, smaller load incre-

ments result in better estimates.

The procedure developed in this study provides a powerful

tool for the evaluation of stresses and deflection of straight, pris-

matic rectangular box girders. Residual stresses can be incorporated,

so can the transverse and longitudinal stiffeners. The effects of the

concrete deck thickness need to be studied further for inelastic

analysis through dividing the deck thickness into layers of finite

elements, a formidable task. Girders with trapezoidal cross-section,

with taper, or with horizontal curvature should be examined. The in-

corporation of geometric nonlinearity also needs to be made to render

the analysis more general. At the present, it appears that a

~67-
parametrix study on the influence of box girder geometry by using the

procedure of this report would provide helpful information for the

design, analysis, and erection of box girders for bridges, buildings,

and other structures.

-68-
 TABLE 2.1 - IN-PLANE STIFFNESS HATRIX

16 i3 D11 12 n12 8 p Dll -12 012 -16 fl D

11 12 012 - 8 f3 Dll -12 012
24 013 16 f3 Dl3 -16 a n33 8 i3 D13 8 C1 n
33 -16 i3 013 -24 n
13 - 8 i3 Dl3
16 C1 n33 16 C1 n -16 C1 n 8 C1 D23 - 8 C1 D33 - 8 ()( 023
23 23
12 n 12 033 -12 033 -12 n33
33
16 C1 o
22 12 012 -16 ()( 022 -12 n12 8 lY 0
22
-12 0
12 - 8 0
22
24 023 8 [>, 013 8 f3 0 -16 fl 0 -16f:lD - 8 i3 013 -24 n
33 13 33 23
16 13 033 -16 C1 n 8 a 0 23 - 8 et D23 - 8 i3 n33
23
-12 OJ:)__ 12 0 -12 D33
33
16 ~ o 11 -12 012 - 8 i3Dll 12 n12 -16 f30u -12 D
12
-24 o13 16 f3 o13 24 o
13 - 8 i3 DlJ. 8 a D33 -16 1:J n13
16 C1 n33 16 C1 o23 - 8 C1 n
33 - 8 ()( 023 8 C1 D23
-12 o33 12 o 12 033
33
16 0' 022 12 0 - 8 ()( 022 12 0 8 ()( 022
12 12
-24 0
23 - 8 i'l o13 24 02':3 -16 f3 D -16 f3 n
13 33
I 16 p 0 33 - 8 ()( 023 - 8 ~ 0 33 8 ()( 0 23
Q'\
\0 12 n -12 0
I
33 33
16 p n -12 0 8 f3 Dll 12 n12
11 12
-24 D 16 f3Dl3 -16 a n 8 f3 o13
13 33
16 C1 n 16 0' 023 -16 0' 023
33
-12 033 -12 033
Symmetric with 16 ()( 022 -12 012 -16 C1 n
22
respect to diagonal -2!1 023 8 f3 Dl3 8 i3 033
16 ~ n33 -16 a n
23
12 n
33
16 ~ n11 12 n
12
a
()( = -b 24 n13 16 i3 013
16 a n 16 0' 023
33
i2 n33
p =~
16 ()' 022
Multiplier 24 n
t 23
48
16 s 033
 TABLE 2.2

COMPARISONS OF TWO DEGREE OF FREEDOM AND FIVE DEGREE OF

.FREEDOM RESULTS

Problem 1 2 3 4 5

Structure. Box Box Box Box Wide Flange

Loading Bending Torsional Bending Torsional Bending

Division Along
the Half 10 10 6 6 10
Length
-
Mesh Coarse Coarse Medium Medium

Figure 2.2b 2.2b 2.2c 2.2c 2.3

Maximum 2 DOF 0.04068 0.04556 0.04098 0.04323 0.07131

Deflection
(in.) 5 DOF 0.03976 0.04454 0.04072 0.04239 0.07126

Maximum 2 DOF -0.2948 -0.3099 -0.2738 -0.2902 -8.1963

Top Flange
Stress 5 DOF -0.2821 -0.2912 -0.2793 -0.2896 -8.1925
(cr ksi)
n

Maximum 2 DOF 7.0549 6.5223 6.6931 6.6776 8.1925

Bottom
Flange 5 DOF 6.7857 6.2105 6.6781 6.6533 8. 1925
Stress
(cr - ksi)
n

CP Time 2 DOF 0.818 1. 644 1. 7139 3.437 0.755

for
Solution 5 DOF 7. 906 15.449 18.822 37.137- 7.872
(sec)

24/164 24/164 38/168 38/168

I~
2 DOF 20/165
h

Total Degrees 5 DOF 55/368 55/368 87/381 87/381 47/396

of Freedom

-70-
 TABLE 2.3

EFFECT OF .MESH DIVISION ALONG THE LENGTH

Deflection at Midspan Stresses at Top and Bottom

(P/Et) Flanges at Midspan (P/ht)

L/2 2h 4h 8h 2h 4h 8h

a/b 1:1 2:1 4:1 1:1 2:1 4:1

-
.,,
FEM 16.35 88.20 344.1 4.667 7.000 7.00

1
Beam:f
Theory 18.19 132.59 1033.29 6.000 12.000 24.00

% 89.9 66.5 33.3 77.7 58.3 29.2

.,•:
FEM 4.37 21.30 120.51 0.8239 1.4465 2.3504

I Beam:f
Theory

%
4.40

99.5
22.20

95.6
154.27

78.1
0.8568

96.2
1. 7136

84.4
3.4272

68.6

·k
FEM 3.034 16.31 88.28 0. 613 1.151 1. 741

Beam:f

D Theory

fo
2.975

102.0
17.95

90.0
131.9

66.9
0.75

81.7
1.5

76.7
'

3.0

58.0

for finite element analysis ~ =0

:fbeam theory including shear deformations

-71-
 TABLE 2.4 EFFECT OF DIVISIONS ACROSS THE FLANGES

Simple Bending Pure Torsion

Maximum Flange Maximum
Number of Divisions Vertical Normal Vertical Horizontal Shear Stresses (ksi )
Deflection~·: Stress f. Deflection Deflection Top Flange Web Bottom Flange
(a) (in.) (b) (ks i) (c) (in.) (d) (in.) (e) (f) (g)

-y-I- .1 -·-
0.03829 5.9464 0.004909 -0.005257 0.02733 0 . 8985 -0.4976
~ -----··· -·

l.-1-~l -- 0.03862 6. 5446 0.004917 -0.005288 0.02418 0.8957 -0.05806

- ------·-·
· r ~ - _:1 0. 03870 6.6476 0.004924 -0.005274 0.02109 0.8123 -0.5826
f--- - --
I
--..1
N
I
0.03870 6. 6792 0.004327 -0.005267 0.02063 0. 7753 -0.5664

1
,• •
11 r
• • I
If

r.. ~:.: _~..l 0.03872 6.6957 0.004961 -0.005234 0.01994 0.7508 -0.6255
1--- --·- ·
- ·· LH1l.!1- ...
: I I . II

..t'!l .l..:...~
l· :!:
}.
l
.
0.03873 6.6992 0.004960 -0.005237 0.01968 0.7344 -0.5963

0.03872 6.7003 0.004932 -0.005260 0.01967 0. 7228 -0.6291

r--- - - ----
0.03876 6.7005 0.004965 -0.005228 0.01988 0. 7161 -0.6116

--L±i, -
~ ' 1 4'· I
.\ (..,

·'
(· L
all stresses and displacements at x = -
2
 TABLE 2.5 EFFECT OF NUMBER OF DIVISIONS ACROSS THE WEB

Simple Bending Pure Torsion

Number of Maximum Maximum
Divisions Across Vertical Bottom Flange Web Shear Vertical Bottom Flange Web Shear
The Web Deflection Normal Stress Stress Deflection Normal Stress Stress
(a) (in.) (b) (ksi) (c) (ksi)

0.03820 6.6847 2.9446 0 , 004590 -0.4784 1.1642.

0.03853 6.5152 2.1290 0~004851 -0.2553 0.5909

I
.......
VJ
I
0.03861 6.5439 2.4340 0.004316 -0.2602 0. 7967

0.03864 6.5487 2.3617 0.004340 -0.2651 0. 7244

0.03866 6.5518 2.4378 0.004352 -0.2673 0.7819

0.03866 6.5536 2.4062 0.004358 -0.2685 0.7532

- --- -~----~ --~ ---- -· --
- ~~~~--------------~----------~----------~------------~---------
i'·
1 ' · .p:n. f t ' 1 F ·J'

-r=r~ -_[4\
 TABLE 2.6

EFFECT OF ELEMENT THICKNESS

Difference
Loading 2 DOF 5 DOF % of 5 DOF

Deflection 1.429 1. 255 13~83

(a)
(in.)
Hulticell Simple Bending
Normal Stress 2.732 1.255 25.72
Concrete (b)
(ksi)
Box Girder
t
Deflection 2.009 1.209 66.10
-....J Pure Torsion
+'
I Normal Stress 3.812 2.085 82.82

E Deflection 1.456 1.449 0.46

n = -s = 10 Simple Bending
E
c Normal Stress 27.811 27.713 0.35
Same Box
Girder with Deflection 2.023 1.975 2.43
Transformed
Pure Torsion
Steel Section Normal Stress 38.586 37.809 2.06
 TABLE 2. 7

WEB SLENDERNESS

D/t 50 100 125 158 200

t (in.) 0.240 0.102 0.096 0.076 0.060

Maximum
Deflection 0.02538 0.03251 0.03532 0.03870 0.04275
(a)

Top Flange
-0.2568 -0.2652 -0.2664 -0.2672 -0.2678
Normal Stress
(b)
-
Bottom Flange
Normal Stress 4.1498 4.9733 5.1882 5.3850 5.5557
(c)

Web Shear
(d)
0.5830 1.2345 l. 5 786 2.0434 2.6527

Shear Flow 0.1399 0.1481 0.1516 0.1553 0.1592

Maximum Web
4.9926 6.0900 6.3934 6.6800 6. 9369
Normal Stress
(e)

Maximum
0.002827 0.003839 0.004316 0.004327 0.005685
Deflection
(a)

Top Flange
0.02709 0.02524 0.02459 0.02393 0.02333
Shear
(b)

Bottom Flange
-0.5058 -0.5281 -0.5360 -0.5448 -0.5546
Shear
(c)

Web Shear
(d)
0.1901 0.4453 0.5851 0. 77 53 1.0236

Shear Flo\v 0.04562 0,5344 0.05617 0.05892 0.06142

-75-
 (a}

(b)

Fig. 2.1 Finite Element Model and Element Coordinate Axes

-76-
 (

JP=IO kips
,Diaphragm AL ~~Dicphragm 7
rr~;::::;:=s:=;=::::;:::========-ir: r, . ,:---~-· - ~~~
L.C ..____.____..______:c_J 1 :JJ
;;;;:;b,~y /J, ?>.t>- I 7;;}T

[.,N Divisions 60" ,~J

( (J )

I
-.....!
-.....!
I

A-/j, /J.-A
(b) (c)

Fig. 2.2 Box Girder Specimen and Mesh Division

 I
-...!
co
I

---~l---
11
,)., .
W8x31
E==29,600 ksi :r
··1.:i

r-----.l.l 'I'
1 - - - _ _ _ _lL o t!..__..9

Fig. 2.3 Wide Flange Beam

 ~
DISTAI\I:E ALOf\G THE LENG-TH I
'
I
12 24 35 48 60

Q)
.01
~ ~
.02

.03 ~
q)
IT
7Tk
Five Degrees of
-
Freedom
Q) I
.04 e> Q)

.05
+ THo Degrees of Freedom

OEFLECTI ON[ IM:HES l ·

'
COMPARISON OF DEFLECTION FOR PROBLEM 1 SIMPLE BENDING

<£
DISTANCE ALONG THE LENG-TH I
I
I
12 24 35 48 60

.01

.02

.03
I
.04 0 Five Degrees of Freedom i
.05 + 1\vo Degrees of Freedom 4l
!
I
DEFLECTIONCINCHESJ
COMPARISON Of DEFLECTIONS FOR PROBLEM 2 TORSIONAL LOADING

Fig. 2.4 Comparison of Deflections from '11:·70 and Five

Degree Freedom Formulation
-79-
 (Symm.)
Cf_
@

l --~----_-------·"
:
.
.~
v
J-.,.0.1
--r. ksi
r

~--~
\
t.JJ\

\
\
\
\
,_ ~
\
+ \
\
\
. __.
~~
I

~~----~_j_ +1.0 ksi

+ 1.0 ksi
~ Five of Freedom
r
+ o Two Degrees of Freedom·
I

L ________ --· ·--n - - - - Tt'0fH'Y

I .....

r n.. ~ ,J I
'\.) u

Fig. 2.5 Normal Stress Distribution for Problem 2

(Cross-section at 50 in. from Left Support)

-80-
 l_P · A

77];7T ===-=--=-=
v]
]:=::!=~-=-~··=::~··=~-=4=-="'=7Mr
Le---_ _ L _ -A_ ~J

lP
tzz.:z:2~~r
l :: : :
- 1.' i
h
IJ
-H,-
I
-r-~~~t
t:-·~--' jL____,
~~-~h, J
A--/\ (a)
<t (Symm.)
a=Lk
P/2 !
1
t~J--)'1 f[J.1 L/2 Ofb
b" h/2 [l~-...,_.;;:·----..---.-..--41 il . -2-h-~--,-:-1-
L~_LI .:~'L/// ·----·+------<".11 h 4h
2=1

L Half Beo._.~i_n_L__:/2 Variab!e _. jl Bh

4 1
'

(b)

Fig. 2.6 Aspect Ratio and Hesh Division Along the Length

-81-
 Stress Deflection (inches)
(ksi)
·+-
Deflection under Web
I
!
+i
I

7.0 TI I Maximum Normal Stress at Bottom Flange

i
i \ . ' ·--·+--- ------- --+------+----
! _.!.-----r
.. ---.
+
..!--'
6,[l .....-; .
_,,- "'
,/

n-,.-, " ""

""
!
5·8 + .: ,j J I
/ "
;

II
!
4·0
+I
·-------------------------
Number of Divisions along the Half Span Length
----- -~

["
- .. ···--· .. -·--·-·-····· . -
.. ) 1G
(2: 1) (1: 1)

"\.
•·· o20 T
'
li (Aspect Ratio)
j
)<.
l
··,25 T
i
i
~ /Maximum Top Flange Stress

~
'
I
!

----------~x~.------~H~--

---30 -~

Stress
(ksi)

Fig. 2. 7 Effect of Mesh Division Along the Length Under Simple

Bending

-82-
 Deflection (inches)
.0049

\Maximum Deflection under Web

.0048

.0047

.0046 Number of Divisions along the Half Span Length

5 10 15

-.5} 2
(2: 1) (1: 1) (3: 1)
(Aspect Ratio)

-.52
-+---+--\-+---+---+--+------+-- ----- -+----
. .+-~
.
-.53 ~.......... Maximum Shear Stress at Bottom Flange
I
I
I
I
I
I
-.54 +

Stress
(ksi)

Fig. 2.8 Effect of Mesh Division Along the Length Under Torsional
Bending

-83-
 ·:;)

--t::=.----~~-·
~ ----~~.. J

/ j~--
~ +------+-------+
,, .....
i'' ....- { - - - - - - -

,, , ... {/2 ~ P/2

7

D
I
I
I
I
,-, ..-., '
• rv,.:l.')<~- +
I
I

I
I
I

10 Division along the Half Span

i I
I

-0383 --~ I
I

+
! I
+ Division across the Web
,Q](;;~
i
+
! 0 Division across the Flange
,Q}'JJ t
j
Number of Divisions across the Web and Flanges
;---+-·-···---+-·-·-·- ··+·--------+ ·---····-·+----·--+-·-·- ·+·-----·+··---··-·
.~~ r:J
!i
.j ·1 r·
l-- • .1 £....:. ._")

22 26 31 36 41 Semiband Width

158 186 214 242 270 Total Degrees of Freedom

0.178 1.081 1.722 2.434 3.366 Solution Time (CP sec.)

Number of Division Increases across the Web

Fig. 2.9 Relative Increase of Accuracy and Computational Time

with Mesh Division

-84-
 . P =I 000 k ~ P=IOOO k
~~ .
1. 2 d lap I~ .. o~··-·
11
tl ji~-~~ i_j../ -~:~ t~
c/ y v
~ rr-

~~-----1------------m~-,~-T------- 1
~ ~r.----3-~_o_'-~.l~;-_3_0_'--:·"" .
1
30 fr -.<:

a. Elevai·ion
1 P= I 000 k
~
~
61/2 II f1
.

b. Cross Section
~ ~p

~ ! : i . i ~---·-·-.:. . . . . . -[·--,i_,---.---.-:----,..---,......!. . .!. . -1

~--_L____J..__.L..l__-ll'----'--'-1-.....l I ! L ~
5.0 ~
1
~?T
1
/ 2@ 2.tl\ 4 @ 5.0 5 4 @ 2.5', 4@ 2 @ 2 . 5'Jm.
Half Bridqo-60'-o"
------·--------
c. iv1esh Division Along The Length

IT-r_j ._____ , J

,.~,::.··· h
",·.:!..:.,:::;,. I Mesh 2
f; ~- ,..,
.--.1 f'\

Fig. 2.10 · Tl·lO Span Multicell Box Girder, Cross-section and Hesh
Sizes

. -85-

.. .,
 ...I A B

-
0
.:w:
DISTANCE ALONG THE LENGTHCFEETl
§. o.ooo 7.500 15-000 22-500 30-000 37.500 45-000 52-500 60-000
0~~~----~--------~------~~------~---------+--------~--------~--------~
-
( /)
w Ip = 1000 k.
:Co
ug
z .
-
..... 0
....
i
A
'iJ

Jm? :A
B
CD
lu
3 p = 1000 k.

~
Q:::C
I
C/) wB
CJ •
X Mesh I I
Ci'
zo
I
I
.::::lC'\1 -;,': Mesh I
t- + Scordelis Mesh I
z
We
z:::g
w.
6 Scordelis Mesh I I
Jo,~
uo ~ Folded Plate Theory
O:CY'J
_J

-
a..
(/)

CJ

Fig. 2.11 Vertical Deflection Along the Longitudinal Line at the Top of the Loaded Web
...•
g
A 0 fSTRNCE l INCHES l B
ij o.ooo 11.200 Z2.400 33.600)(10
0

X = 17.5 ft.
~
0 .....

X Mesh I
~ 0
~(
Mesh I I
"" + Scordelis Mesh I

6 Scordelis Hesh I I
~ (l) Folded Plate Theory
~

j X
I

= 17.5 I
+
X = 30'
II

n#m·
A B

I
-
0

e
0
A
o.ooo
OISTANCEliNCHESl
n.wo :z:z .400 .
B
33.600)10
1

~ X 30 ft.
...
0

Fig. 2.12 Vertical Deflection of Bottom Flange at Transverse

Sections

-87-
 -
t!P
I 1000 k.
0
..... Normal Stress at the Corner
:><
0
= 1000 k.
8 I

I I I
0

I
(0

0
8•
0
(V'J

22.500 30-000 3?.500 '45-000 52.500

l
C/)
C/)
I
-
t5
Zo
a:g
_l,
U..o
(V'J
z'
......
X Mesh I
l/)
0 ;'~
Mesh II
.
(/)
lU 8
~ 0
1- (0 + Scordelis Mesh I
(/) I
/::;. Scordelis Mesh I I
0 Folded Plate Theory
0
8•
0
01
Fig. 2.13 Longitudinal Distribution of Normal Stress (cr ) Along a Line in Top Flange
I n
 ~
I
10 20 30 40 60

ID Beam Theory

lOC.O + Finite Element Analysis

200-0
p
X
E CENTER LlNE DEFLECTION OF W8X31

l
10 20 30 40 60

Beam Theory

-40 + Finite Element

Analysis

.so
.so
l-00
x p FLANGE NORMAL STRESS OF H 6X31

Fig. 2.14 Stress and Deflection of W8 X 31

-89-
 .0__c_m_.________________~~
____2_2_1__

-----
-!il-0.304 em 0.405 emf

at::

~~---------~·----~L~~~-~~-------~~~
v.::>em
0.612 ~ em ·"""· em ---· _
7'8.5 ~ 78.5

c=~.J (a)

L.11'.6--r ~\
~

. .
I

''"'"'0
c.. "() ... 0 ':> 0
10610 1,;.) c...... ~,0
v. 3.0 ':!Q
~-+ I I -t--+---1 -~----; .::>. -+--;,;-·---r-1__;.;_.---;
0.8 1.2 .
6 7D8 10 I! 12 !~) 14 15 16 1"7 18 19
<t-~;·:r.~~.(~.:)-·~~-·ft......__,-0··-~-c-- ---t;J-·- - - r &
~l 5 21 Division A!c)ng Ths Half Spl1n Length
.
1 Sem .1 B ·v 1..1,.! ~1 'n, .• ··-
Jan c~' \~-' /) a.....

J:
:1

(b)
t·.!o ' De,...,;''c\c..
• ~ \j I \_• '(., of Freedom :924

6.25 em~
--t11.0 6.25 em.
~----~------~~------,~,,:-:----t~~~-~--
6.25'· em. 6.25 ern. 1
"""1
5 6
r~r:
1
c
a[:'}9-----...-----;s:-----~y
9 1oj
0 I

~4
!6 Division / 1.long The Half Span Length 1

Two Degree Five Degree

\ ·'I'd'
;'". t
v I il i ! ' 29
•'"'•"'\.
I
68
r\J o. [1'::.:·11-r:•t:•
.• ,_ ... ':J v,.. , 0:: 1 F•·o::.dom:
,.,u.._ ~74
1 '-.. 884 .l ..._ •• (c)

Fig. 2.15 Cross-section and Finite Element Models of a Thin Flange

-90-
 B=

/
78.5 ern

+
78.5/ -/1, . -

+
1..00
L

-75

Mmax
-60 _
0 =- --
h.t.i3 8 Test Results

h = 6. 69 ern
+ Finite Elements
Schmidt's Prediction

-25

I
A
I
.
~- --~----~:------·---~-----}--·~~-------+, --- ----11 B
+i--
O.OJJ S.OOJ lO.OJO JS,COO 2G~OOO

Fig. 2.16 Top Flange Normal Stresses at a Transverse Section

-91-
 an /
-a

1-25

t.otre

-75

-50 I
M
max i
(J= h. t.B Hesh I 1\vo Degrees of Freedom !
h = 6.69 em + Mesh I I Five Degrees of Freedom I
-25 0 Mesh I I 1\vo Degrees of Freedom I
I
I
A !B
o.oon s.ooo 10.000 15.000 20.000 25.000
I
DISTANCE FROM THE WEB
I
I
Fig. 2.17 Effect of ~vo Degree versus Five Degree Freedom
Formulation on a Thin Flange

-92-
 /3/16 11
x 1" ( t yp.)

1-.-c-~·~-r1!
,J_l __ --1-J-=---~-1-,J -- - II
p
-;~-

L.,.
.....

Elevation

---·
!"'""~:,-------·
3 3 il
-----·

11
12

Se\,
,...·.·~--
II u:l :--'1\-. ·-;-',.
•,

Fig. 2.18 Dimensions and Geometry of Girder Dl

-93-
 DISTANCE ALONG THE LENGTH
40 60 80 100 120

.075

.150
LOAOlNG Dl-CB-C37 !LOAD P=24.Q KIPSl
(a)

- Finite Element Analysis

6. Thin \.Jalled Elastic Beam Theory

If)
El Test Results
I
0

9
0
0
N
I

80
0~--~--+-~----+-------+-------+-------~----~
1
2.4oo 4.aoo 1 .eoo 9.600 14 •.JOIDHO
DISTANCE ALONG THE LENGTH
B
0
0
N a

LOADING Dl-C0-617 !LOAD P=ll·O ~IPSl

(b)

Fig. 2.19 Deflection Along the Length of Girder Dl

 -r Finite Element Analysis
Thin Halled Elastic Beam Theory
0 Test Results

+ +
f l

Scale

1 in. 5 ksi
+

STRESS OISTRIBUTI(}l AT X::00-25 DUE TO. 16-0 KIP OF LOOO EC= 7-668

Fig. 2.20a Normal Stress Distribution due to a Torsional Loading

on Girder Dl (86.25 in. from Left Support)

-95-
 Finite Element Analysis

Thin Walled Elastic Beam Theory

~ Test Results

• + + iQ + + ..

+ .
+ ' \~
I ------- --- __:~
- :.51..-----------'

Scale

1.0 in. 7. 5 ks i j
+

STRESS 015TRIBUT10N AT X=56-25 JN. DUE TO 18. KIP OF LOAD ,EC=D-0

Fig, 2.20b Normal Stress Distribution due to Simple Bending of

Girder Dl (56,25 in, from Left Support)

-96-
 Deflection (inches)
(Stress (ksi))

0.02 ip +p 5.0 k
(4.0) M.,.
~

.!IT I :ht
7m"r.
(3.0)
\
MT 5.0 X 15 = 75.0 k-ft.
~
+-.......
...... _..L

0.01
f",,
-
........ ...,. .....

+--- --------
(2.0)
--- ----
I
\0 - ---+-----------
'-....]
I
1-Jeb Shear at Midspan -------+----~--------- ------+---
0.005
( 1. 0) ~~--------------------~~----------------~------------------0---·
Deflection under Web

(inches)

0.25 0.50 0.75 1.00

Thickness of the Diaphragm
--------------
Fig. 2.21 Effect of Diaphragm Thickness on Stresses and Deflection under Pure
Torsional Loading
 lp::: 1.0 kips
~ .

l}ymm
I

-- --I I .

·~ _J_· : -~
, :-~~-~--
r ~---·""
·1
I
_tL,,[:r--
;
~ . _j'l
{~
I I

L------ ~ -------
. I
L,,..../' I V
I I -~;~- I -r
~- - - ~--- --- -
·1 crn.=O.I ksi
---
Tension

Fig" 2.22 Principal Stresses in a Diaphragm

-98-
 OISTRNCE ALONG THE LENGTHCINCHESl ~
I
6 12 18 24 30 36 42 48 54 50

.010 Diaphragm Locations

0 0
"'k L/1
.020 I 0 L/2

I 6 L/3
ctI
MT =j75 . 0 k~ in .
I + L/4
. 030
4+ I I I X L/5
~ ::;;;9?
I
\.0
\.0 I I I I I I 0 L/6
~
.p = ~
I

L
.040 ( 5.0 kip

.050

.060 ~
OISPlRCEHENTC[NCHESJ
Fig. 2.23 Effect of Diaphragm Spacing on Deflection, Load at Mid-Span
 DISTANCE ALONG THE LEHGTHClNCHESJ I
A 6 12 18 24 30 36 42 4B 54 50

B!
IMT i
F\ -1+- I
.002 ~ IB ~
I
~ L/2.
It
1 I

.003 Diaphragm Locations

{!) l.c/2

.005 A L/3
I
t-'
0
0
+ L/4
I

.006 X L/s

I I ~ L/b
.ooa

.009
OISPLACEMENTClNCHESl

Fig, 2.24 Effect of Diaphragm Spacing on Deflection, Load at Midspan,

Exaggerated Scale.
 (

DISTANCE ALONG THE LENGTHCINCHE5l

6 12 lB 24 30 36 42 46 54

.006

Diaphragm Locations Diaphragm Spacing ,·

.010 D 0

L/1
i
I
.015 0 L/2

6 L/3

I
.020 I; I -:·· L/4
...... L
0
<t- ."'
}-'
MT MT I I X L/5
I
i
.025 i
~ +7
11TSm I I I I I <> L/6
mfbT.
.030 fP ~ p = 5.0 kip

.035 I II A
7777777m

.040
D15FLRCEMENTCINCHE5l

Fig. 2.25 Effect of Diaphragm Spacing on Deflection, Load

at Quarter Points
 OISTRNCE ALONG THE LEHGTHClNCHESJ
6 12 lB 24 30 36 42 48 54 60

0 L/2
.002 L/3

+ L/4

x L/5
.002
I I I I I L/6

.003
I
,.._.
0
N
I

.004 M,. l'v\i

?'<= ~
~ ?7,9m
j~ ~I
L
.005

.oos
OISPLRCEHENTCINCHESJ
Fig. 2.26 Effect of Diaphragm Spacing on Deflection, Load at Quarter Points,
Exaggerated Scale
 Stress (tension)
·O""u ----

E ( s ·;rain).

- cry
Stress (compress ion)

(a)

Stress (compression)
Equ. 3.1
f I
c I
I
0.7fc
I I
I
I
I
I
Strain ·
Tension Compression Eu

Stress (tension)
(b)

Fig. 3.1 Stress-Strain Relationship for Steel and Concrete

·-103-
 ()2
era

Curve., ---- ....... '\

/
/ " \
~,.....__~!-Initial Yield Curve
I I
I
I
I 1.0 I Oj
I // o-o
I /

__ _
\ /
- /
\ -/ VonMises Yield Curve
........ /

'
Cracked _CJ!_I =- 15 Crad~ed
/ o-2
-._.. ___ ..__,
A .,

Cracked

Crushed

Fig. 3.2 Yield Criteria for Steel and Concrete

-lOL;.-
 P (force)

Incrementa I Solution

(displacement)

LJ
~8J.
8

Fig. 3.3 Graphical Representation of Incremental Tangent

Stiffness Procedure

-105-
 p p

vV8x40

J .. -'·~
11 11
56 56 56"

11
8.06
~ ~

11 11
t w = 0.370 8.32

11

>-fct
11
56 28
..... 1 Eil .

~ ~p · .Symm.
-~"!II I y I
Ir· -· .I .,

~ .
·~·nit i -i •· -·...
~
I
I

. . t% .
o.5 .., o.1s·~ o.as .., o.90 11, J.o", 2.0 .., 3.5 11, 3@ 12 11, 2®a'~ 12", s ..
.
1..
5@ 0.7768 = 3.884" .
11
f'

_.___-=.-::_ ..J~"----e---j
l.~i·o1~'l ~.o1gj
Fig. 3.4 Loading and Finite Element Discretization of
a Steel I-Beam

-106-
 120
FORCECKIPSJ

100
0 Test Results from Ref. 67

+ Analytical Solution
80

I
60
>--'
0
-....J
I

20

.4 .s .a 1.0 1.2 1.4 1.6

DEFLECTIONClNCHESl
MID SPAN DEFLECTION OF B X 40 WIDE FLANGE BERM
Fig. 3.5 Load Deflection Curve for a Steel I-Beam
 Ml

I@ t@ t@ I® I® Jill w
1.. j ... 511 . . I.. 411 .... I 3 .. _,.., 41;2 3 4 y2
II _ , -z• II"' ,_ II ..-I Ill
11
26
A_J
11

r 7 (MI)
-
311 J, II
1--+---- 1 64

Section A- A

Fig. 3. 6 Dimensions ~nd Geometry of Specimens- Nl and H2

-108-
 LOAOlKfPSl

0 Test results
6 Analytical, no diaphragm
+ Diaphragm is included, fine load
step

1.0 0 Diaphragm is included, coarse

load step

~MT-:2P
!PJMT

[ I]
~
I I I
~.
l-1
12"
;j'
12" J
.o4 .oa .12 .16 .20
OEFLECTIO~rr~HESl
Loqo VERSUS ~TOSPAN OEFLECTfON Of Ml(UNSYM. LOAD'
Figc 3.7 Load versus Hidspan Deflection- Specimen Hl

-109-
 LORO PfKIPSl

3.0

Analytical Results

""Test Results

First Signs of Noticeable Web

Bulging in Panel 5

p
1.0
~
p

+
5
N
.,.. J.
5" 3"

a.o 12.0 20.0

SHEAR STRESSCKSil

Fig. 3.8 Load versus Shear Stress at Centerline of Panel 5 -

Specimen M2

-llO
ao.o
FORCE£KlPSl

70.0

60.0

0 Test Results from Ref. 28

50.0 + Analytical Solution with Fine Mesh
6 Analytical Solution with Coarse
Mesh

40.{}

30.0

P/2 p'/2
~ ~
20.0
t /

£ ~ :1 I I
~
k--~
10.0
60" I

.50 }.00 1.50 2a00 2.50 3.0{} 3.50 4 .. 00

OEFLECTION£INCHESl

Fig. 3.9 Load versus Midspan Deflection - Girder Dl

-111-
60
FORCEfKJPSl

50

G Test Results
+ Theoretical Curve

30

20 p
p

~ ~
I 16 1 ~
I
10
l
1
60"
~
1 6~

.50 1.00 1.50 2.00 2.50 3a00 3.50

DEFLECTIONrrNCHESl

Fig. 3,10 Load versus Hidspan Deflection- Girder D2

-112-
80
LORD PCKIP5l

71

B Test Results

62 + Analytical Solution

63

44

36
6011 60 11
f 1
k
f
27 Jp

19
£2 X =
I
I
56.25
31
J/2 (/2 0

u
6 11 17 23 28 34 39 45
NORMAL STRESSCK5Il

Fig. 3.11 Load versus Normal Stress at 3.75 in. away from
Midspan at the Bottom Flange
-113-
 I

r
I

P=56.89 kips

I d 5.0" 5.0
11
2.5
11
2.5
11

Scale=
0
' '.0 .0
2
ksi + Yielded Elements

Fig. 3.12 Principal Stresses in the Web Adjacent to Lbading

Point

-llL~-
 P/2 P/2

1
11
60

-l- -t- -t- -t - + -+-- + -;- - +- - + -+- -+ yj

+~-;-T-+-+-+-+-+-+-~-~~
t+ -+-T--+ -+ --r-t-+-+-t --9
+-+-4 -+ -+ -+ -+- -+ -+ --+

P=70.1 kips
122L:J Yielded ~ Strain Hardening Started

Fig. 3.13 Yield Zones of the Web of Girder Dl under Different ·

Load Magnitudes !
-ll5-
 REFERENCES

1. ASCE-AASHO Committee on Flexural Members_, Subcommittee on Box

Girder Bridges, ''Trends in the Design of Box Girder Bridges",
Journal of the Structural Division, ASCE, Vol. 93, No. ST3,
Proc. Paper 5278, June 1967, pp. 165-180

2. ASCE-AASHO Task Committee on Flexural Members, Subcommittee

oN Box Girders, Progress Report on Steel Box Girder Bridges",
Journal of the Structural Division, ASCE, Vol. 97, No. ST4,
Proc. Paper 8068, April 1971, pp. 1175-1185

3. Culver, Charles G., "Steel-Box Girder Bridges Ultimate Strength

Considerations", Journal of the Structural Division, ASCE,
Vol. 100, No. ST12, Proc. Paper 11014, December 1974,
pp. 2433-2448

4. Rockey, K. C., Bannister, J. L., Evans, H. R., editors,

"Developments in Bridge Design and Construction, Crosby
Lock';·lOOd and Son, Ltd, London 1971

5. International Association for Bridge and Structural

Engineering, "Reports of the Working Commissions", Seminar on
"Design of Plate and Box Girders for Ultimate Strength", Vol.
11, London, 1971

6. "Inquiry Into The Basis of Design and Method of Erection of

Steel Box Girder Bridges", Report of the Committee, Appendix I,
Department of Environment, Scottish Development Department,
Welsh Office

7. Cartledge, P. (Editor), "Steel Box Girder Bridges", Proceedings

of the International Conference Organized by the Institution
of Civil Engineers in London, February 13-14, 1973

8. Petzold, E. H., Galambos, T. V., "Behavior and Desigri of Large

Steel Box Girder Bridges", Washington University, Civil and
Environmental Engineering Department Research Report No. 26,
Structural Division, December 1973

9. Goldberg, J. R., Leve, H. L., "Theory of Prismatic Folded Plate

Structures", Publications IABSE, Vol. 17, 1957

10. Scordelis, A. C., "Analysis of Simply Supported Box Girder

Bridges", Report No. SESM 66-17, Department of Civil Engineering,
University of California, Berkeley, California, 1966

-116-
11. Johnston, S. B. a:nd Mattock, Ao H., "Lateral Load Distribution of
Load in Composite Box Girder Bridges", Highway Research Record,
No. 167, High\vay Research Board, Washington, D. C., 1967

12. Fountain, R. S. and Mattock, A. H., "Composite Steel-Concrete

Multicell Box Girder Bridges", Proceedings of Canadian
Structural Engineering Conference, Toronto, Canada, February
1968

13. DeFries-Skene, A. and Scordelis, A. C., "Direct Stiffness

Solution for Folded Plates", Journal of the Structural Division,
ASCE, Vol. 90, No. ST4, Proc. Paper 3394, August 1964

14. Vlasov, V. z., "Thin Walled Elastic Beams", United States

Department of Commerce, PST Catalog No. 428, 1959

15. Dabrowski, R., "Curved Thin-Walled Girders, Theory and Analysis",

Translation, Cement and Concrete Association, No. 144, London,
1972

16. Wright, R. N., Abdel-Samed, S. R., Robinson, A. R., "Analysis

and Design of Closed-Section Girder Bridges with Diaphragms",
Final Report submitted to the American Iron and Steel Institute,
Project 110, March 1967

17. Abdel-Samed, So R., Wright, R.N., Robinson, A. R., "Analysis

of Box Girders with Diaphragms", Journal of the Structural
Division, ASCE, Vol. 94, No. STlO, Proc. Paper 6153, October
1968

18. Wright, R. N., Abdel-Samed, S. R., Robinson, A. R., "BEF Analogy

for Analysis of Box Girders", Journal of the Structural Division,
ASCE, Vol. 94, No. ST7, Proc. Paper 6025, July 1968

19. Sisodiya, R. G. and Ghali, Amin, "Analysis of Box Girder Bridges

of Arbitrary Shape", International Association for Bridge and
Structural Engineering Publication, Vol. 33-I, 1973

20. Scordelis, A. C., "Analysis of Continuous Box Girder Bridges",

Report No. SESM 67-25, Department of Civil Engineering, University
of California, Berkeley, 1967

21. Sawka, F. and Cope, R. J., "Analysis of Multi-cell Bridges With-

out Transverse Diaphragms -A Finite Element Approach", The
Structural Engineer, London No. 11, Vol. 47, November 1969

22. William, K. J. and Scordelis, A. C., "Computer Program for

Cellular Structures of Arbitrary Plain and Geometry", Report
No. UC-SESM 70-10, University of California, Berkeley, 1970

-ll7-
23. Criesfield, M. A., "Finite Element Methods for the Analysis of
Multi Cellular St~uctures", Proceedings of Institution of Civil
Engineers, London, Vol. 48, 1971

24. Cheung, Yo K., "Analysis of Box Girder Bridges by Finite Strip

Method", Paper presented at 2nd International Symposium on
Concrete Bridge Design, Chicago, March 1969

25. Parr, D., "An Ultimate Strength Analysis of Box Girder Highway
Bridges", Ph.D. Dissertation, New Mexico State University, May
1968

26. Corrado, J. Ao, "Ultimate Strength of Single-Span Rectangular

Box Girders", Ph.D. Dissertation, Lehigh University, 1971

27. Corrado, J. A. and Yen, B. T., "Failure Tests of Rectangular

Model Steel Box Girders", Journal of the Structural Division,
ASCE, Vol. 99, No. ST7, Ju~y 1973

28. Yen, B. T., et al, "Tests on Model Composite Box Girders'' (Draft),
Fritz Engineering Laboratory Report No. 380.6, Lehigh University,
October 1973

29. Zienkie~;vicz, 0. C., "The Finite Element Method in Engineering

Science", McGraw Hill Book Company, London, 1971

30. Desai, C. S. and Abel, J. F., "Introduction to Finite Element

Method", Van Nostrand-Reinhold Company, New York, 1972

31. Roland, I. and Bell, K. (Editors), "Finite Element Hethod in

Stress Analysis", Tapir For lag, Trondheim, Nonvay, 1969

32. Pian, T. H. H., "Derivation of Element Stiffness Matrices by

Assumed Stress Distribution", AIAA Journal, Vol. 2, No. 7,
July 1964, p. 931

33. Fraeys de Veubeke, B., "A Conforming Finite Element for Plate
Bending", International Journal of Solids and Structures, Vol.
4, No. 1, January 1968, p. 95

34. Herrmann, L. R., "A Bending Analysis for Plates", Proceedings

of the Matrix Methods in Structural Mechanics, Wright Patterson
AFB, Ohio, AFFDL-TR 66-80, November 1975, p. 577

35. Arderheggen, E., "Finite Element Plate Bending Equilibrium

Analysis", Journal of the Engineering Hechanics Division, ASCE,
Vol. 95, No. EM4, August 1969, p. 841

36. Clough, R. H. and Tocher, J. L., "Finite Element Stiffness

Matrices for Analysis of Plate Bending", Proceedings of
Conference on Matrix Hethods in Structural Mechanics, Wright
Patterson AFB, Ohio, AFFDL-TR 66-80, 1965, p. 515
-118-
37. Wegmuller, A. W. and Kostem, C. N., "Finite Element Analysis
of Plates and Eccentrically Stiffened Plates", Fritz Engineering
Laboratory Report No. 378A.3, Lehigh University, Bethlehem, Pa.
February 1973

38. Clough, R. W. and Felippa, C. A., "A Refined Quadrilateral

Element for Analysis of Plate Bending", Proceedings of the
Second Conference on Marrix Methods in Structural Mechanics,.
Wright Patterson AFB, AFFDL-TR 68-150, October 1968, p. 399

39. Abel, J. F. and Desai, C. S., "Comparison of Finite Elements for

Plate Bending", Journal of the Structural Division, ASCE, Vol.
98, No. ST9, September 1972

40. Tocher, J. L. and Hartz, B. J., "Higher Order Finite Element

for Plane Stress", Proceedings ASCE, Journal of Engineering
Mechanics Division, EM4, August 1967

41. Clough, R. W., "The Finite Element Method in Structural

Mechanics", Chapter 7 of "Stress Analysis" edited by Po C.
Zienkiewicz and G. S. Hollister, John Wiley and Sons, New York,
1965

42. Wegmuller, Ao Wo, "Finite Element Analysis of Elastic-Plastic

Plates and Eccentrically Stiffened Plates", Ph.D. Dissertation,
Lehigh University, 1971

43. Peterson, H. S., Kostem, C. N. and Kulicki, J. M., "Inelastic

Analysis of Reinforced Concrete Slabs", Fritz Engineering
Laboratory Report No. 378B.3, Lehigh University, Bethlehem, Pa.,
May 1974

44. Yilmaz, C., "Documentation for Program BOXGIR11 , Fritz

Engineering Laboratory Report, Lehigh University, Bethlehem, Pa.,
(in preparation)

45. Schmidt, H., "Die Mittragende Wirkung Der Fahrbahnen Breiter

Plattenbaikem-Brucken", The Flange Performance of Wide Bridge
Decks, Ph.D. Dissertation, Carolo-Wilhelmina Technical University,
Brunschweig, 1970

46. Kulicki, J. M. and Kostem, C. N., "The Inelastic Analy-sis of

Prestressed and Reinforced Concrete Bridge Beams by the Finite
Element Method'', Fritz Engineering Laboratory Report Noo. 378A. 6,
September 1973

47. Lin, C. S., "Nonlinear.Analysis of Reinforced Concrete Slabs

and Shells", Report No. SESM 73.7, Department of Civil
Engineering, University of California, Berkeley, 1973

48. Ostapenko, A., Chern, C. and Parsanejad, S., "Ultimate Strength

Design of Plate Girders", Proceedings of the Conference on
"Developments in Bridge Design and Construction", Cardiff, 1971
.-119-
49. Rockey, K. C., Evans, H. R. and Porter, D. M., "The Ultimate
Load Capacity of ·stiffened Webs Subjected to Shear and Bending",
International Conference on Steel Box Girder Bridges, The
Institution of Civil Engineers, Westminster, London,
February 13-14, 1973

50. Przemieniecki, J. S., "Theory of Matrix Structural Analysis",

McGraw-Hill Book Company, New York, 1968

51. Amen, H. and Levine, H. S., "Plasticity-Theory and Finite

Element Applications, Advances in Computational Methods in
Structural Mechanics and Design", edited by J. T. Oden, R. W.
Clough and Y. Yamamoto, UAH Press, 1972

52. Meldenson, A., "Plasticity: Theory and Application", McMillan

Company, New York, 1968

53. Kupfer, H., Hilsdorf, H. K. and Rusch, H., "Behavior of Concrete

Under Biaxial Stresses", ACI Journal, Vol. 68, August 1969

54. Kupfer, H. and Gerstle, K. H., "Behavior of Concrete Under

Biaxial Stresses", Journal of Engineering, Mechanics Division,
ASCE, Vol. 99, No. EM4, August 1973, pp. 853-866

55. Calladine, C. R., "Engineering Plasticity", Pergamon Press,

London, 1969

56. Yamada, Y., Yoshimura, N. and Sakurai, T., "Plastic Stress

Strain Matrix and Its Application for the Solution of Elastic-
Plastic Problems by the Finite Element Method", International
Journal of Mechanical Sciences, Pergamon Press, Vol. 10, 1968,
pp. 343-354

57. Yamada, Y., "Recent Japanese Developments in Matrix Displacement

Method For Elastic Plastic Problems", presented at the Japan-
United States Seminar on Matrix Methods of Structural Analysis
and Design, Tokyo, Japan, August 25-30, 1969

58. Zienkiewicz, 0. C., Valliappan, S. and King, I, P., "Elasto-

Plastic Solutions of Engineering Problems, Initial Stress
Finite Element Approach", International Journal of Numerical
Methods in Engineering, Vol. I, 1969

59. Meldelson, A. and Manson, S. S., "Practical Solutions of Plastic

Deformation Problems in Elastic-Plastic Range", NACA TN 4088,
September 1959

60. Lansing, W., Jensen, W. and Falby, W., "Matrix Analysis Methods
for Inelastic Structures", Proc. of First Air Force Conference
on Matrix Methods in Structural Mechanics, AFFDL, TR 66-80,
November, 1965, pp. 605~634

-120-
61. Pope, G. G., "The Application of the Matrix Displacement Method
In Plane Elasto-Plastic Stress Problems", Proc. of First Air
Force Conference on ~~trix Methods in Structural Mechanics,
AFFDL, TR 66-80, November 196 , p. 635

62. Marcal, P. V. and King, I. P., ''Elastic Plastic Analysis of Two

Dimensional Stress Systems by the Finite Element Method 11 ,
International Journal of Mechanical Science, Vol. 9, 1967

63. Struik, J. H. A., "Applications of Finite Element Analysis to

Non-Linear Plane Stress Problems 11 , Ph.D. Diss.ertation,
Department of Civil Engineering, Lehigh University, Bethlehem,
Pa., 1972

64. Oden, J. T., ''Finite Element Application in Nonlinear

Structural Analysis 11 , Proceedings of the Symposium on
Application of Finite Element Methods in Civil Engineering,
Vanderbilt University, Nas~ville, Tenn., November 1969

65. Oden, J. T., "Finite Elements of Non-Linear Continua",

McGraw Hill Book Company, New York, 1972

66. Conte, S.D., "Elementary Numerical Analysis", McGraw Hill

Book Company, 1965

67. Knudsen, K. E., Yang, C. H., Johnston, B. G. and Beedle, L. S.,

11
Plastic Strength and Deflections of Continuous Beams", The
Welding Journal Research Supplement, May 1953

-121-
 NOMENClATURE

a half length of a rectangular element

b half width of a rectangular element

[BJ matrix relating nodal point displacements to element

strains

B semi-band width of the total stiffness matrix

compliance matrix

elasticity matrix

elastic-plastic compliance matrix

E Young's modulus of elasticity

E Young's modulus of elasticity for concrete

c
Est strain hardening modulus

F loading function

compressive strength of concrete

tensile strength of concrete

H functional representing effective stress versus

effective strain

H' slope of the effective stress-strain curve

K yield condition

[KJ total global stiffness matrix

[k
e
J element stiffness matrix

element stiffness matrix in global coordinates

in-plane element stiffness matrix

plate bending element stiffness matrix

span length of a girder

122-
m,n constants in Ramberg Osgood equation

N total number of degree of freedom

[N] shape function

N. ,N. individual terms in the shape function

~ J
[P} load vector for the complete system

[l!.P} load increment vector

scaling factor

r. scaling factor for the individual elements

~

r scaling factor that 'tvill be used in the computations

m

[T] transformation matrix

t thickness of an element

[u} element displacement vector

u, v displacement of a point in x and y-direction,

respectively

u. ,u. generalized nodal point displacements

~ J
w displacement in z-direction

constants

Yxy
shearing strain

[ 6} nodal point displacement for the complete system

E: 'E: strain in x-direction y-direction respectively

X y
[e} strain vector

[ei} incremental strain vector

[e } initial strain state

0

rE: }
\. p equivalent plastic strain

E: equivalent plastic strain

p

-123-
(ep} plastic strain vector

(f.:P} plastic strain increment vector

d€p x'd€ p y plastic strain increments

CJ ,CJ ,CJ normal stresses

X y
'T shearing stress
xy
da ,dcr ,dT increments of stresses
x y xy
[a} stress vector

(6-} incremr ~t of stress vector

( CJ '} deviatoric stress vector

£;. 'J increment of deviatoric stress vector

( CJ } initial stress vector

0

CJ effective stress
e
yield stress level

principal stresses in two perpendicular directions

increment in effective stress due to plastic stress

increment

rotation about x, y, z axis respectively

\) Poisson's ratio

dA. a constant in elastic-plastic compliance matrix

-124-
 VITA

The author was born in Uluborlu, Turkey on June 20, 1946,

the second child of Mustafa and Hafize Yilrnaz.

He was graduated from Izrnir Ataturk high school in June

1963. He attended Middle East Technical University in Ankara,

Turkey from September 1963 to June 1968, supported by a ''Kennedy

Scholarship", and received th~ B. S. degree in Civil Engineering.

He carne to Lehigh University in September 1968 under a

fellowship through the Institute of International Education and

completed his M. S. Degree in Civil Engineering in October 1969. He

worked as a graduate assistant at Middle East Technical University

until corning back to Lehigh University in September 1971. He is

supported by the Agency for International Development and involved

with the project dealing with strength of composite box girders.

,. if

He is married to the former Fusun Ozdarnar and has one

daughter, Ebru. The author will return to Turkey and join the staff

of Middle East Technical University in Ankara.

-125-